Decentralized Cooperative Online Estimation With Random Observation Matrices, Communication Graphs and Time Delays

Jiexiang Wang; Tao Li; Xiwei Zhang

REVIEW 4 major objections 4 minor 45 references

Decentralized Cooperative Online Estimation With Random Observation Matrices, Communication Graphs and Time Delays

T0 review · 4 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A network of estimators can converge on a shared unknown parameter even when observation matrices, communication graphs, and communication delays are all random, time-varying, and statistically dependent, provided a stochastic…

desk verdict A genuine theoretical extension of consensus-plus-innovations convergence under a block-wise stochastic persistence-of-excitation condition, with random delays included; the main theorems are conditional and the delayed-case equivalence and numerical verification need tightening, but this deserves peer review. read the letter →

arxiv 1908.08245 v6 pith:IUYGVHOV submitted 2019-08-22 eess.SP cs.LGcs.SYeess.SY

Jiexiang Wang , Tao Li , Xiwei Zhang This is my paper · ORCID

classification eess.SPcs.LGcs.SYeess.SY MSC 93E1093E1593A1460G42

keywords decentralizedonlineestimationcooperativerandomcommunicationgraphobservationmatrixtimedelaypersistenceofexcitationconsensusplusinnovationsmeansquareconvergence

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proves that a group of networked estimators can cooperatively track an unknown parameter even when each node's measurements, the communication links, and the delivery delays are all randomly time-varying and can even be statistically dependent. The central result is a condition called stochastic spatio-temporal persistence of excitation: over each fixed-length block, the pooled conditional expectation of graph Laplacians and observation matrices must have a minimum eigenvalue bounded below by a non-summable sequence. When that condition holds, properly chosen vanishing gains drive all nodes' estimates to the true parameter in mean square and, in the delay-free case, almost surely. This matters because existing analyses typically required i.i.d. graphs, balanced mean graphs, or independence between sensing and communication. The paper's condition allows unbalanced, stepwise-disconnected graphs and correlated random environments, and it handles random nonuniform delays by introducing delay matrices and reducing the delayed system to an equivalent delay-free one.

What carries the argument

The load-bearing object is the stochastic spatio-temporal persistence-of-excitation condition: for a fixed block length $h$, the minimum eigenvalue of the sum of conditional expectations of $b(k)$ times the symmetrized graph Laplacian and $a(k)$ times the Gram observation matrix over each block is required to be at least $c(m)$, with $\sum_m c(m)=\infty$. This condition controls the binomial expansion of the random matrix product $\Phi_P(k,0)$, forcing the expected error covariance to decay through products of factors $(1-c(m)+\text{small remainder})$. For the delayed case, the companion machinery is the delay matrix $I(k,q)$ and the equivalent delay-free system $r(k+1)=F(k)r(k)+g(k)$, whose invertibility is guaranteed by a small initial consensus gain (Lemma V.1); the same blockwise persistence condition then controls the products $\Phi_F(k,0)$.

What would settle it

Run the delay-free algorithm on a Markovian switching chain whose stationary graph is balanced with a spanning tree and whose stationary Gram matrix satisfies (14), set gains like $a(k)=b(k)=1/(k+1)^{0.6}$, and compute the averaged relative error after a long horizon; Corollary IV.1 predicts the error approaches zero, so a nonzero limiting error would refute the central claim. Equivalently, construct a process satisfying all assumptions except (b.1) by making nodes isolated and setting all observation matrices to zero: the PE condition fails and the estimates remain frozen, confirming that the condition is indispensable.

Watch

Extended reading notes

Core claim

The paper's central claim is that consensus-plus-innovations estimation over a random network converges whenever the random environment is persistently exciting in a joint spatio-temporal sense. Concretely, for a block length $h$, the key condition requires that the minimum eigenvalue of $\sum_{k=mh}^{(m+1)h-1}(b(k)E[\hat{L}_G(k)|\mathcal{F}(mh-1)]\otimes I_n + a(k)E[H^T(k)H(k)|\mathcal{F}(mh-1)])$ be at least $c(m)$ almost surely, with $\sum_m c(m)=\infty$. Under this condition, Theorem IV.1 gives mean-square and almost-sure convergence for the delay-free algorithm, and Theorem IV.2 gives the same under a constant lower bound when the digraphs are conditionally balanced. For Markovian switching environments, Corollary IV.1 derives the condition from a stationary balanced graph with a spanning tree plus spatio-temporal joint observability, so neither local observability nor instantaneous global observability is needed. With delays, the paper models the random lags with delay matrices $I(k,q)$ and, through binomial expansion of random matrix products, converts convergence of the delayed error system into convergence of expectations of products of transformed matrices $F(k)$; Corollary V.3 states that under the same persistence-of-excitation condition, any bounded delay can be tolerated by choosing gains sufficiently small when the digraphs are conditionally balanced.

Load-bearing premise

The load-bearing premise is the stochastic spatio-temporal persistence-of-excitation condition itself: over every block the pooled conditional expectations of graph Laplacians and observation matrices must have a minimum eigenvalue bounded below by a non-summable sequence, on every sample path, and the paper's other assumptions do not guarantee this.

Editorial extensions

If this is right

Observation matrices, communication graphs, and delays need not be mutually independent or spatio-temporally independent for convergence; all three may be correlated.
Balanced mean graphs are not necessary: the persistence condition can hold even when sample paths and mean graphs are unbalanced.
For Markovian switching environments, convergence follows when the stationary graph is balanced with a spanning tree and the measurement model is spatio-temporally jointly observable, so neither local observability nor instantaneous global observability is required.
With conditionally balanced digraphs, any bounded random delay can be accommodated by choosing sufficiently small algorithm gains, and the maximum tolerable delay bound is expressed in terms of graph weights, delay probabilities, and gains.
In the limit of completely isolated nodes, the condition degenerates to independent stochastic persistence-of-excitation conditions for centralized estimation.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The condition could be monitored online by computing running blockwise minimum eigenvalues; when it fails, a network could adaptively increase block length or switch to more informative measurements to restore convergence.
The dependence of the admissible delay on graph intensity suggests a protocol-level design rule: if delay distributions are heavy, links must be used more frequently or observation matrices must be richer.
A natural extension is to replace the almost-sure blockwise lower bound with a high-probability version; such a relaxation might give practical stopping rules while preserving convergence in probability rather than almost surely.
The blockwise condition could also be used to compare sensor scheduling policies, since any policy that keeps the blockwise minimum eigenvalue non-summable is sufficient for convergence.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Two things you should know. First, this is a real theoretical advance: it proves mean-square and almost-sure convergence for consensus-plus-innovations estimation under a block-wise stochastic spatio-temporal persistence-of-excitation condition, without requiring observation matrices and graphs to be i.i.d., Markovian, or mutually independent, and it introduces delay matrices to handle random time-varying delays. Second, the main theorems are conditional on that PE condition, which is assumed for general environments and only verified for the Markovian case; and the delayed-case equivalence is asserted without proof. I recommend peer review, not desk reject, but with requested revisions.

What is actually new: the relaxation of independence assumptions; the delay-matrix formulation; the use of binomial expansion of random matrix products to turn convergence analysis into analysis of expectations of matrix products; and Corollary IV.1, which gives a nice sufficient condition — balanced stationary graph with spanning tree plus spatio-temporal joint observability, without requiring local or instantaneous global observability. The appendix proof is long but the structure is coherent: block conditional expectations, product decay from a non-summable c(m), and martingale arguments. The self-citations to [36] and [41] are used as lemmas; that is not a problem.

Soft spots, in order of seriousness. (1) The stochastic spatio-temporal PE condition is not derived for general environments. It is a pathwise lower bound on each length-h block of conditional expectations. For a stationary ergodic process with occasionally poorly excited blocks, it can fail. The theorems honestly state it as a condition, so this is not circular, but the claimed generality — random graphs and observation matrices without special statistical properties — is narrower than it appears: only the Markovian corollary gives a verifiable sufficient case. The paper should say this explicitly. (2) The transformation from the delayed error system (10) to the augmented system (15)-(16) is asserted as 'it can be verified' but no proof is given. All delayed-case results rest on that equivalence; either prove it or give a precise citation. (3) Numerical checks: I believe the λ_min in Remark 5 is about 0.483, not 0.5821, and the values f = 0.0005 and the 0.01 delay factor in Section VI are not consistent with their own formulas (I get roughly 0.000245 and 0.00407). If my arithmetic is right, the figures need to be redone. Minor, but it undercuts the claim that the conditions are verified.

Who this is for: people working on distributed estimation, consensus algorithms, or adaptive networked control who need convergence guarantees under correlated random topologies and delays. They get a serious, if conditional, convergence theory.

My recommendation: send to peer review. The core mathematical structure is sound and the result is a genuine extension. Request fixes for the numerical slips, a proof or reference for the equivalence, and a clearer statement that the general PE condition is assumed, with the Markovian case as the main verifiable path.

Referee Report

4 major / 4 minor

Summary. The paper studies decentralized cooperative online estimation of a fixed unknown parameter by a network of nodes, where each node takes linear measurements with random, possibly correlated observation matrices, and communicates over random directed graphs with random bounded time-varying delays. The algorithm combines a consensus term using delayed neighbor estimates with a local innovation term. The authors derive conditions under which all node estimates converge to the true parameter in mean square and almost surely (delay-free case) or in mean square (delayed case). The key tool is a stochastic spatio-temporal persistence-of-excitation condition on block sums of conditional expectations of graph Laplacians and observation matrices. For Markovian switching graphs and observation matrices, this condition is shown to follow from balancedness with a spanning tree of the stationary graph plus spatio-temporal joint observability. For the delayed case, the error system is transformed into an augmented system involving an auxiliary filter, and convergence conditions are expressed through conditional expectations of delay matrices, graph matrices, and observation matrices.

Significance. If the theorems are correct, the paper makes a useful contribution: it relaxes the usual i.i.d. or Markovian independence assumptions on graphs and observation matrices, allows unbalanced and time-varying mean graphs, and treats nonuniform random delays through a novel delay-matrix formalism. The proof strategy via binomial expansion of random matrix products is sophisticated and appears to be largely self-contained. The Markovian corollary provides a genuinely verifiable sufficient condition and clarifies that neither local observability nor instantaneous global observability is necessary. However, the advertised generality is tempered by the fact that the central stochastic spatio-temporal PE condition is assumed, not certified, for general non-Markovian environments, and the delayed-case analysis rests on an equivalence that is asserted but not proved.

major comments (4)

[Section V, Eqs. (15)-(16)] The equivalence between the delayed error system (10) and the augmented system (15)-(16) is asserted with 'It can be verified' but no proof is given. This equivalence is load-bearing: the proof of Theorem V.1 and all delayed-case corollaries rely on r(k) being equal to e(k). I request a complete derivation or a precise reference that establishes this equivalence, including the construction of F(k) and C_q(k) for all k and the treatment of the initial conditions r(k)=e(k) for -d<=k<=-1.
[Section IV, conditions (b.1) and (c.1)] The stochastic spatio-temporal persistence of excitation condition is a pathwise requirement on every history and is not implied by Assumptions A1-A2 and Conditions C1. The paper supplies a verifiable sufficient case only in the Markovian setting (Corollary IV.1). For general non-Markovian environments, the main theorems are conditional on a premise that is not derived and may fail for stationary ergodic processes that occasionally have long blocks with H(k)=0 and disconnected graphs. The authors should either provide verifiable sufficient conditions for (b.1)/(c.1) in the non-Markovian case or explicitly state that the general theorems apply only when this condition can be certified, and adjust the abstract and introduction accordingly.
[Remark 5] The numerical claim lambda_min = 0.5821 in the two-node example is incorrect. For a12=1, a21=0.3, H1=0, H2=1, the matrix Lhat + H^T H equals [[1, -0.65], [-0.65, 1.3]], whose minimum eigenvalue is approximately 0.4829, not 0.5821. The illustrative claim that the condition holds with c(m)=0.5821/(m+1) should be corrected and re-verified.
[Section VI] The numerical verification of Corollary V.1 contains arithmetic errors. With C1=1, beta_a=1, beta_H=4.07, N=4, d=4 and psi_2=0.01, the function f_{C1,beta_a,beta_H,N,d}(0.01) evaluates to about 2.45e-4, not 5e-4. Similarly, the weighted delay factor sum 0.5 * sum_{q=0}^4 p_q * ((1+psi_2)^q - 1)/(2 - (1+psi_2)^q) is approximately 0.00823, not 0.01. These values are used to set b(0) and to verify inequality (21); the example should be recomputed with correct values and the plots re-examined.

minor comments (4)

[Throughout] There are several typographical errors: 'applicaitons', 'Conditon', 'SCDA system' (should be SCADA), and 'sigma-filed' for 'sigma-field'. A careful proofread is recommended.
[Figure 5 caption] The caption of Figure 5 uses 'Lambda^2_m - 0.01(b(2m-1)+b(2m))' while the text and the y-axis label use 'Lambda^2_m - 0.01 * sum_{k=2m}^{2m+1} b(k)'. These should be made consistent.
[Section II.B, Eq. (3)] The notation p_{ji,q}(k) is used for the probability that the delay equals q, but the delay variable is written as lambda_{ji}(k). Clarifying the indexing (link from j to i) in a sentence would help the reader avoid confusion with the q-th delay matrix.
[Section IV, Theorem IV.2] The definition of the set Gamma_1 requires E[A_G(k)|F(k-1)] to be nonnegative and its associated random graph balanced almost surely. It would be helpful to state explicitly that this is a condition on the conditional mean graph, not on the sample graph, in the main text preceding Theorem IV.2.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: convergence theorems are explicit sufficiency results under an assumed spatio-temporal persistence-of-excitation condition; only a minor technical self-citation appears in the proofs.

full rationale

The paper's central claims are conditional theorems. Theorems IV.1 and IV.2 state that if the stochastic spatio-temporal persistence-of-excitation condition (b.1) or (c.1) holds, then the estimators converge in mean square and almost surely. This condition is an explicit assumption on the environment--it involves conditional expectations of the Laplacians and observation matrices over blocks--and is not derived from, or fitted to, the estimation errors. The proof of Theorem IV.1 analyzes the random product Phi_P(k,0) and shows that the block-product norm decays because c(m) is non-summable with b^2(mh)=o(c(m)); no parameter is fitted to force convergence. Corollary IV.1 derives the PE condition from more primitive Markov-chain assumptions (uniform ergodicity, balanced stationary graph with spanning tree, and spatio-temporal joint observability), using Lemma A.7 and continuity of lambda_min. The delayed-case results similarly rest on explicit conditions such as (20), (21), and (22), with proofs bounding the inverse-matrix terms; these are sufficiency statements, not predictions obtained by fitting. The numerical example chooses a permissible c(m) to illustrate that the condition can hold, which is allowed by the existential quantifier in the theorems and is not a fitted-input-called-prediction. A limitation, but not circularity, is that for general non-Markovian environments the PE condition is assumed rather than guaranteed by A1-A2; only Corollary IV.1 supplies primitive conditions under which it holds. The only self-citation in the proof machinery is 'By Lemma A.1 in [36]' in the almost-sure part of Theorem IV.1; it is a technical conditional-expectation/martingale lemma used in passing and is not load-bearing for the main convergence claim. Hence no circular step is exhibited, and the score of 2 reflects only that minor, non-load-bearing self-citation.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The central claim is conditional on a persistence of excitation condition and on boundedness or martingale assumptions. No free parameters are fitted to data. The delay matrices are a bookkeeping device, not a new physical entity. The main unproved step is the equivalence between the delayed system and the augmented system (15)-(16).

assumptions (6)

domain assumption Measurement noise v(k) is a martingale difference sequence with uniformly bounded conditional second moment (Assumption A1.b).
Used throughout the proof to zero out cross terms involving noise and to bound noise contributions.
domain assumption Observation matrices and adjacency matrices are uniformly bounded a.s. (Assumption A2.b).
Needed for the matrix product estimates in Lemmas B.1 and C.1.
domain assumption Stochastic spatio-temporal persistence of excitation: the minimum eigenvalue of the block sum of conditional expectations of b(k)LG_hat(k) tensor I plus a(k)H^T(k)H(k) is bounded below by c(m) with the sum of c(m) divergent (conditions b.1 and c.1).
This is the key environment condition; the convergence theorems are conditional on it. It is not derived from primitive assumptions except in the Markovian corollary.
domain assumption Conditionally balanced digraphs: E[A_G(k)|F(k-1)] is nonnegative and its associated graph is balanced a.s. for all k (set Gamma_1).
Required for the delay-case corollaries to ensure the symmetrized expected Laplacian is positive semidefinite.
ad hoc to paper Equivalence of time-delay system (15)-(16) with the original error system (10) is asserted with 'It can be verified' and not proved.
All delayed-case results rely on this equivalence, yet no derivation is included.
standard math Standard martingale convergence and conditional moment inequalities (Robbins-Siegmund, conditional Lyapunov, conditional Holder).
Used in the stochastic convergence proofs; standard references are cited.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Decentralized Cooperative Online Estimation With Random Observation Matrices, Communication Graphs and Time Delays." pith.science (2026). https://pith.science/paper/IUYGVHOV

@misc{pith2026190808245,
  author       = {Pith},
  title        = {Pith review of: Decentralized Cooperative Online Estimation With Random Observation Matrices, Communication Graphs and Time Delays},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IUYGVHOV}},
  note         = {Machine review of arXiv:1908.08245}
}

read the original abstract

We analyze convergence of decentralized cooperative online estimation algorithms by a network of multiple nodes via information exchanging in an uncertain environment. Each node has a linear observation of an unknown parameter with randomly time-varying observation matrices. The underlying communication network is modeled by a sequence of random digraphs and is subjected to nonuniform random time-varying delays in channels. Each node runs an online estimation algorithm consisting of a consensus term taking a weighted sum of its own estimate and neighbours' delayed estimates, and an innovation term processing its own new measurement at each time step. By stochastic time-varying system, martingale convergence theories and the binomial expansion of random matrix products, we transform the convergence analysis of the algorithm into that of the mathematical expectation of random matrix products. Firstly, for the delay-free case, we show that the algorithm gains can be designed properly such that all nodes' estimates converge to the true parameter in mean square and almost surely if the observation matrices and communication graphs satisfy the stochastic spatiotemporal persistence of excitation condition. Secondly, for the case with time delays, we introduce delay matrices to model the random time-varying communication delays between nodes. It is shown that under the stochastic spatio-temporal persistence of excitation condition, for any given boundeddelays, proper algorithm gains can be designed to guarantee mean square convergence for the case with conditionally balanced digraphs.

Figures

Figures reproduced from arXiv: 1908.08245 by the authors.

**Figure 1.** The curves of f1,1,1,N,d(·). If the conditions of Lemma V.1 hold, then F(k) is invertible a.s. Thus, by (16), we have F(k) = INn − b(k)DG(k) ⊗ In − a(k)HT (k)H(k) + b(k)A(k, 0) − C1(k) = INn − b(k)DG(k) ⊗ In − a(k)HT (k)H(k) + b(k)A(k, 0) − (C2(k) − b(k)A(k, 1))F −1 (k − 1) . . . = INn − G(k), k ≥ 0, (17) where G(k) , b(k)DG(k) ⊗ In + a(k)HT (k)H(k) − b(k) X d q=0 A(k, q) [PITH_FULL_IMAGE:figures/full_fig_p014_1.png] view at source ↗

**Figure 2.** IEEE-14 multi-area buses and the communication graphs. [PITH_FULL_IMAGE:figures/full_fig_p018_2.png] view at source ↗

**Figure 3.** Curves of Λ 2 m and c(m) = 0.0112 (2m+2)0.52 w.r.t. m. 0 2000 4000 6000 8000 10000 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 P4 i=1 kxi(k) − x0k/(4kx0k) k Case with link and sensing failures Case without link and sensing failures [PITH_FULL_IMAGE:figures/full_fig_p019_3.png] view at source ↗

Figures from the paper (1 more)

**Figure 5.** Figure 5 [PITH_FULL_IMAGE:figures/full_fig_p019_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 44 canonical work pages

[36]

Distributed averaging with random network graphs and noises,

T. Li and J. Wang, “Distributed averaging with random network graphs and noises,” IEEE Trans. Information Theory , vol. 64, no. 11, pp. 7063-7080, Nov. 2018

work page 2018
[41]

Guo, Time-V arying Stochastic Systems: Stability and Adaptive Theory (Second Edition)

L. Guo, Time-V arying Stochastic Systems: Stability and Adaptive Theory (Second Edition). Beijing, China: Science Press, 2020

work page 2020
[1]

Abur and A

A. Abur and A. G. Exposito, Power System State Estimation: Theory and Implementation. Boca Raton, FL, USA: CRC Press, 2004

work page 2004
[2]

Y . B. Shalom, X. R. Li, and T. Kirubarajan, Estimation With Applications to Tracking and Navigation. New York, USA: Wiley, 2001

work page 2001
[3]

Covariance intersection algorithm for distributed spacecraft state estimation

P. O. Arambel, C. Rago and R. K. Mehra, “Covariance intersection algorithm for distributed spacecraft state estimation”, in Proc. Amer . Contr . Conf., Arlington, V A, USA, 25-27 Jun. 2001, pp. 4398-4403

work page 2001
[4]

N. E. Leonard, and A. Olshevsky, Cooperative learning in multiagent systems from intermittent measurements, SIAM. J. Control and Optimization , vol. 53, no. 1, pp. 1-29, 2015

work page 2015
[5]

A distributed state estimation approach to condition monitoring of nonlinear electric power systems,

G. Rigatos, P. Siano, and N. Zervos, “A distributed state estimation approach to condition monitoring of nonlinear electric power systems,” Asian J. Control , vol. 15, no. 3, pp. 1-12, Jul. 2012

work page 2012
[6]

Parallel and distributed state estimation

D. M. Falcao, F. F. Wu and L. Murphy, “Parallel and distributed state estimation”, IEEE Trans. Power Systems , vol. 10, no. 2, pp. 724-730, May 1995

work page 1995

Show all 45 references

[7]

Distributed LMS for consensus-based in-network adaptive processing,

I. Schizas, G. Mateos and G. Giannakis, “Distributed LMS for consensus-based in-network adaptive processing,” IEEE Trans. Signal Processing , vol. 57, no. 6, pp. 2365-2382, Jun. 2009. December 8, 2020 DRAFT 47

2009
[8]

Consensus+innovations distributed Kalman ﬁlter with optimized gains,

S. Das and J. M. F. Moura, “Consensus+innovations distributed Kalman ﬁlter with optimized gains,” IEEE Trans. Signal Processing, vol. 65, no. 2, pp. 467-481, Jan. 2017

2017
[9]

Optimal recursive estimation with uncertain observation,

N. E. Nahi, “Optimal recursive estimation with uncertain observation,” IEEE Trans. Information Theory , vol. 15, no. 4, pp. 457-462, Jul. 1969

1969
[10]

Distributed robust estimation over randomly switching networks using H∞ consensus,

V . Ugrinovskii, “Distributed robust estimation over randomly switching networks using H∞ consensus,” Automatica, vol. 49, no. 1, pp. 160-168, 2013

2013
[11]

Gossip and distributed Kalman ﬁltering: Weak consensus under weak detectability,

S. Kar and J. M. F. Moura, “Gossip and distributed Kalman ﬁltering: Weak consensus under weak detectability,” IEEE Trans. Signal Processing , vol. 59, no. 4, pp. 1766-1786, Apr 2011

2011
[12]

CIRFE : A distributed random ﬁelds estimator,

A. K. Sahu, D. Jakoveti ´c and S. Kar, “CIRFE : A distributed random ﬁelds estimator,” IEEE Trans. Signal Processing, vol. 66, no. 18, pp. 4980-4995, Sep. 2018

2018
[13]

FADE: Fast and asymptotically efﬁcient distributed estimator for dynamic networks,

A. Sim ˜oes and J. Xavier, “FADE: Fast and asymptotically efﬁcient distributed estimator for dynamic networks,” IEEE Trans. Signal Processing , vol. 567, no. 8, pp. 2080-2092, Apr. 2019

2019
[14]

Diffusion least-mean squares over adaptive networks: Formulation and performance analysis,

C. G. Lopes and A. H. Sayed, “Diffusion least-mean squares over adaptive networks: Formulation and performance analysis,” IEEE Trans. Signal Processing , vol. 56, no. 7, pp. 3122-3136, Jul. 2008

2008
[15]

Diffusion LMS strategies for distributed estimation,

F. S. Cattivelli and A. H. Sayed, “Diffusion LMS strategies for distributed estimation,” IEEE Trans. Signal Processing , vol. 58, no. 3, pp. 1035-1048, Mar. 2010

2010
[16]

Robust distributed estimation by networked agents,

S. Al-Sayed, A. M. Zoubir and A. H. Sayed, “Robust distributed estimation by networked agents,” IEEE Trans. Signal Processing, vol. 65, no. 15, pp. 3909-3921, Aug. 2017

2017
[17]

Diffusion estimation over cooperative multi-agent networks with missing data,

M. R. Gholami, M. Jansson, E. G. Str ¨om and A. H. Sayed, “Diffusion estimation over cooperative multi-agent networks with missing data,” IEEE Trans. Signal and Information Processing over Networks, vol. 2, no. 3, pp. 276-289, Sep. 2016

2016
[18]

Diffusion adaptation over multi-agent networks with wireless link impairments,

R. Abdolee, B. Champagne and A. H. Sayed, “Diffusion adaptation over multi-agent networks with wireless link impairments,” IEEE Trans. Mobile Computing , vol. 15, no. 6, pp. 1362-1376, Jun. 2016

2016
[19]

Diffusion LMS with correlated regressors I: Realization-Wise stability,

M. J. Piggott and V . Solo, “Diffusion LMS with correlated regressors I: Realization-Wise stability,” IEEE Trans. Signal Processing, vol. 64, no. 21, pp. 5473-5484, Nov. 2016

2016
[20]

Diffusion LMS with correlated regressors II: Performance,

M. J. Piggott and V . Solo, “Diffusion LMS with correlated regressors II: Performance,” IEEE Trans. Signal Processing , vol. 65, no. 15, pp. 3934-3947, Aug. 2017

2017
[21]

Diffusion LMS ﬁlter for distributed estimation of systems with stochastic state transition and observation matrices,

J. Y . Ishihara and S. A. Alghunaim, “Diffusion LMS ﬁlter for distributed estimation of systems with stochastic state transition and observation matrices,” in Proc. Amer . Contr . Conf.,Seattle, W A, USA, 24-26 May 2017, pp. 5199-5204,

2017
[22]

Distributed parameter estimation in sensor networks: Nonlinear observation models and imperfect communication,

S. Kar, J. M. F. Moura and K. Ramanan, “Distributed parameter estimation in sensor networks: Nonlinear observation models and imperfect communication,” IEEE Trans. Information Theory , vol. 58, no. 6, pp. 3575-3605, Jun. 2012

2012
[23]

Consensus+innovations distributed inference over networks: Cooperation and sensing in networked systems,

S. Kar and J. M. F. Moura, “Consensus+innovations distributed inference over networks: Cooperation and sensing in networked systems,” IEEE Signal Processing Magazine , vol. 30, no. 3, pp. 99-109, May 2013

2013
[24]

Distributed parameter estimation over unreliable networks with markovian switching topologies,

Q. Zhang and J. F. Zhang, “Distributed parameter estimation over unreliable networks with markovian switching topologies,” IEEE Trans. Automatic Control , vol. 57, no. 10, pp. 2545-2560, Oct. 2012

2012
[25]

Distributed ﬁltering over wireless sensor networks with parameter and topology uncertainties,

J. Zhang, X. He and D. Zhou, “Distributed ﬁltering over wireless sensor networks with parameter and topology uncertainties,” International J. Control, DOI: 10.1080/00207179.2018.1489146, 2018

2018
[26]

Mahmoud, Robust Control and Filtering for Time-Delay Systems

M. Mahmoud, Robust Control and Filtering for Time-Delay Systems . New York, USA: Marcel Dekker, 2000

2000
[27]

Delay-distribution-dependent load frequency control of power systems with probabilistic interval delays,

C. Peng and J. Zhang, “Delay-distribution-dependent load frequency control of power systems with probabilistic interval delays,” IEEE Trans. Power Systems , vol. 31, no. 4, pp. 3309-3317, Jul. 2016

2016
[28]

Time synchronization in WSNs with random bounded communication delays,

Y . P. Tian, “Time synchronization in WSNs with random bounded communication delays, ” IEEE Trans. Automatic Control, vol. 62, no. 10, pp. 5445-5450, Oct. 2012

2012
[29]

Leader-following-based distributed Kalman ﬁltering in sensor networks with communication delay,

Y . Zhang, F. Li and Y . Chen, “Leader-following-based distributed Kalman ﬁltering in sensor networks with communication delay,” J. the Franklin Institute , vol. 354, no. 16, pp. 7504-7520, Sep. 2017. December 8, 2020 DRAFT 48

2017
[30]

Distributed consensus-based estimation considering network induced delays and dropouts,

P. Mill ´an, L. Orihuela, C. Vivas and F. R. Rubio, “Distributed consensus-based estimation considering network induced delays and dropouts,” Automatica, vol. 48, no. 10, pp. 2726-2729, Jul. 2012

2012
[31]

Consensus for linear multiagent systems with time varying delays: A frequency domain perspective,

Y . Chen and Y . Shi, “Consensus for linear multiagent systems with time varying delays: A frequency domain perspective,” IEEE Trans. Cybernetics , vol. 47, no. 8, pp. 2143-2150, Aug. 2017

2017
[32]

Distributed consensus for multiagent systems with communication delays and limited data rate,

S. Liu, T. Li, and L. Xie, “Distributed consensus for multiagent systems with communication delays and limited data rate,” SIAM J. Control and Optimization , vol. 49, no. 6, pp. 2239-2262, Aug. 2011

2011
[33]

Distributed consensus for multi-agent systems with delays and noises in transmission channels,

S. Liu, L. Xie and H. Zhang, “Distributed consensus for multi-agent systems with delays and noises in transmission channels,” Automatica, vol. 47, no. 5, pp. 920-934, Mar. 2011

2011
[34]

A. J. Wood and B. F. Wollenberg, Power Generation, Operation, and Control . New York, NY , USA: Wiley, 2012

2012
[35]

Laplacians and the Cheeger inequality for directed graphs,

F. Chung, “Laplacians and the Cheeger inequality for directed graphs,” Annals of Combinatorics , vol. 9, no. 1, pp. 1-19, 2005

2005
[37]

Consensus problems in networks of agents with switching topology and time- delays,

R. Olfati-Saber and R. M. Murray, “Consensus problems in networks of agents with switching topology and time- delays,” IEEE Trans. Automatic Control , vol. 49, no. 9, pp. 1520-1533, Sep. 2004

2004
[38]

Estimating time-varying parameters by the Kalman-ﬁlter based algorithm,

L. Guo, “Estimating time-varying parameters by the Kalman-ﬁlter based algorithm,” IEEE Trans. Automatic Control , vol. 35, no. 2, pp. 141-147, Feb. 1990

1990
[39]

S. P. Meyn and R. L. Tweedie, Markov Chains and Stochastic Stability . London, UK: Springer-Verlag, 1993

1993
[40]

Zhou and J

K. Zhou and J. C. Doyle, Essentials of Robust Control . Upper Saddle River, NJ, USA: Prentice-Hall, 1998

1998
[42]

A convergence theorem for nonnegative almost supermartingales and some applications,

H. Robbins and D. Siegmund, “A convergence theorem for nonnegative almost supermartingales and some applications,” In Selected Papers, T. L. Lai, and D. Siegmund, Eds. New York, NY , USA: Springer-Verlag, 1985

1985
[43]

Kallenberg, F oundations of Modern Probability, 2nd ed

O. Kallenberg, F oundations of Modern Probability, 2nd ed. New York, NY , USA: Springer-Verlag, 2002

2002
[44]

Closed-loop control of systems over communications network with queues,

H. Chan and U. Ozguner, “Closed-loop control of systems over communications network with queues,” International Journal of Control , vol. 62, no. 3, pp. 493-510, 1995

1995
[45]

A new method for stabilization of networked control systems with random delays,

L. Q. Zhang, Y . Shi, T. W. Chen, and B. Huang, “A new method for stabilization of networked control systems with random delays,” IEEE Trans. Automatic Control , vol. 50, no. 8, pp. 1177-1181, Aug. 2005. December 8, 2020 DRAFT

2005

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [36]

Distributed averaging with random network graphs and noises,

T. Li and J. Wang, “Distributed averaging with random network graphs and noises,” IEEE Trans. Information Theory , vol. 64, no. 11, pp. 7063-7080, Nov. 2018

work page 2018

[2] [41]

Guo, Time-V arying Stochastic Systems: Stability and Adaptive Theory (Second Edition)

L. Guo, Time-V arying Stochastic Systems: Stability and Adaptive Theory (Second Edition). Beijing, China: Science Press, 2020

work page 2020

[3] [1]

Abur and A

A. Abur and A. G. Exposito, Power System State Estimation: Theory and Implementation. Boca Raton, FL, USA: CRC Press, 2004

work page 2004

[4] [2]

Y . B. Shalom, X. R. Li, and T. Kirubarajan, Estimation With Applications to Tracking and Navigation. New York, USA: Wiley, 2001

work page 2001

[5] [3]

Covariance intersection algorithm for distributed spacecraft state estimation

P. O. Arambel, C. Rago and R. K. Mehra, “Covariance intersection algorithm for distributed spacecraft state estimation”, in Proc. Amer . Contr . Conf., Arlington, V A, USA, 25-27 Jun. 2001, pp. 4398-4403

work page 2001

[6] [4]

N. E. Leonard, and A. Olshevsky, Cooperative learning in multiagent systems from intermittent measurements, SIAM. J. Control and Optimization , vol. 53, no. 1, pp. 1-29, 2015

work page 2015

[7] [5]

A distributed state estimation approach to condition monitoring of nonlinear electric power systems,

G. Rigatos, P. Siano, and N. Zervos, “A distributed state estimation approach to condition monitoring of nonlinear electric power systems,” Asian J. Control , vol. 15, no. 3, pp. 1-12, Jul. 2012

work page 2012

[8] [6]

Parallel and distributed state estimation

D. M. Falcao, F. F. Wu and L. Murphy, “Parallel and distributed state estimation”, IEEE Trans. Power Systems , vol. 10, no. 2, pp. 724-730, May 1995

work page 1995