Abstract
A subspace-based deterministic identification method is proposed for MIMO linear systems with time delay in their state variables. This method can estimate the matrices of the state-delayed system, its order and the amount of time delay. First, by extending the orthogonal decomposition subspace method for state-delayed systems, the augmented system matrices which include the original system matrices are identified. Then, an algorithm is presented to find a similarity transformation that converts the augmented matrices to a special form through which matrices of the state-delayed system can be extracted. The values of the time delay and model order are estimated by introducing appropriate criteria. Finally, three examples are provided to validate the proposed identification method.
Similar content being viewed by others
References
I.B. Abdelwahed, A. Mbarek, K. Bouzrara, Adaptive MPC based on MIMO ARX-Laguerre model. ISA Trans. 67, 330–347 (2017)
H. Akaike, A new look at the statistical model identification. IEEE Trans. Autom. Control 19, 716–723 (1974)
J.J. Batzel, H.T. Tran, Stability of the human respiratory control system I. Analysis of a two-dimensional delay state-space model. J. Math. Biol. 41, 45–79 (2000)
L. Belkoura, M. Dambrine, Y. Orlov et al., Identifiability and identification of linear systems with delays, in Advances in Time-Delay Systems, ed. by S.I. Niculescu, K. Gu (Springer, Berlin, 2004), pp. 123–135
L. Belkoura, Y. Orlov, Identifiability analysis of linear delay-differential systems. IMA J. Math. Control Inf. 19, 73–81 (2002)
E.K. Boukas, Z.K. Liu, Deterministic and Stochastic Time Delay Systems (Springer, Boston, 2002)
D. Di Ruscio, Combined deterministic and stochastic system identification and realization: SDR—a subspace approach based on observations. Model. Identif. Control 17(3), 193–230 (1996)
F. Ding, Y. Gu, Performance analysis of the auxiliary model-based stochastic gradient parameter estimation algorithm for state-space systems with one-step state delay. Circuits Syst. Signal Process. 32, 585–599 (2013)
B. Ding, L. Xie, C. Wenjian, Robust MPC for polytopic uncertain systems with time-varying delays. Int. J. Control 31(8), 1239–1252 (2008)
W. Favoreel, B. De Moor, P. Van Overschee, Subspace state space system identification for industrial processes. J. Process. Control 10, 149–155 (2000)
Y. Gu, F. Ding, Auxiliary model based least squares identification method for a state space-model with a unit time-delay. Appl. Math. Model. 36, 5773–5779 (2012)
Y. Gu, F. Ding, Parameter estimation for a multivariable state space system with d-step state-delay. J. Frankl. Inst. 350, 724–736 (2013)
Y. Gu, F. Ding, J. Li, State filtering and parameter estimation for linear systems with d-step state-delay. IET Signal Process. 8(6), 639–646 (2014)
Y. Gu, R. Ding, A least squares identification algorithm for a state space model with multi-state delays. Appl. Math. Lett. 26, 748–753 (2013)
K. Gu, V.L. Kharitonov, J. Chen, Stability of Time-Delay Systems (Springer, Boston, 2003)
Y. Gu, X. Lu, R. Ding, Parameter and state estimation algorithm for a state space model with a one unit state delay. Circuits Syst. Signal Process. 32(5), 2267–2280 (2013)
T. Katayama, Subspace Methods for System Identification (Springer, Heidelberg, 2005)
T. Katayama, H. Ase, Linear approximation and identification of MIMO Wiener–Hammerstein systems. Automatica 71, 118–124 (2016)
S.M. Kissler, C. Cichowitz, S. Sankaranarayanan et al., Determination of personalized diabetes treatment plans using a two-delay model. J. Theor. Biol. 359, 101–111 (2014)
V. Kolmanovskii, A. Myshkis, Introduction to the Theory and Applications of Functional Differential Equations (Springer, Dordrecht, 1999)
W.E. Larimore, Canonical variate analysis in identification, filtering and adaptive control, in: Proceedings of the 29th Conference on Decision and Control, pp. 596–604, 1990
B. Lehman, E. Verriest, Stability of a continuous stirred reactor with delay in the recycle stream, in: Proceedings of the 30th Conference on Decision and Control, pp. 1875–1876, 1991
L. Ljung, System Identification: Theory for the User (Prentice-Hall, Englewood Cliffs, 1995)
V. Lunel, M. Sjoerd, Identification problems in functional differential equations, in: Proceedings of the 36th IEEE Conference on Decision and Control, pp. 4409–4413, 1997
Y. Orlov, L. Belkoura, J.P. Richard et al., Identifiability analysis of linear time-delay systems, in: Proceedings of the 40th IEEE Conference on Decision and Control, pp. 4776–4781, 2001
Y. Orlov, L. Belkoura, J.P. Richard et al., On identifiability of linear time-delay systems. IEEE Trans. Autom. Control 47, 1319–1324 (2002)
Y. Orlov, L. Belkoura, J.P. Richard et al., Online parameter identification of linear time-delay systems, in: Proceedings of the 41st IEEE Conference on Decision and Control, pp. 630–635, 2002
Y. Orlov, L. Belkoura, J.P. Richard et al., Adaptive identification of linear time-delay systems. Int. J. Robust Nonlinear Control 13, 857–872 (2003)
Y. Orlov, I.V. Kolmanovsky, O. Gomez, On-line identification of SISO linear time-delay systems from output measurements: theory and applications to engine transient fuel identification, in: Proceedings of the 2006 American Control Conference, pp. 1179–1184, 2006
J.H. Park, S. Han, C.G. Park et al., On-line model parameter estimations for time-delay systems, in: Proceedings of the 7th Asian Control Conference, pp. 626–629, 2009
G. Picci, T. Katayama, Stochastic realization with exogenous inputs and ‘subspace-methods’ identification. Signal Process. 52, 145–160 (1996)
C.D. Schaper, W.E. Larimore, D.E. Sborg et al., Identification of chemical processes using canonical variate analysis. Comput. Chem. Eng. 18(1), 55–69 (1994)
G.W. Stewart, Perturbation theory for the singular value decomposition. UMIACS-TR-90-124 (1990)
A.J. Van der Veen, E.F. Deprettere, A.L. Swindlehurst, Subspace based signal analysis using singular value decomposition. Proc. IEEE 81, 1277–1308 (1993)
P. Van Overschee, B. De Moor, N4SID: subspace algorithms for the identification of combined deterministic–stochastic systems. Automatica 30, 75–93 (1994)
P. Van Overschee, B. De Moor, Subspace identification for linear systems: theory, implementation, applications (Kluwer Academic Publishers, Norwell, 1996)
M. Verhaegen, Identification of the deterministic part of MIMO state-space models given in innovations form from input–output data. Automatica 30, 61–74 (1994)
M. Verhaegen, P. Dewilde, Subspace model identification part 1. The output-error state-space model identification class of algorithms. Int. J. Control 56(5), 1187–1210 (1992)
J. Wang, S.J. Qin, A new subspace identification approach based on principal component analysis. J. Process. Control 12, 841–855 (2002)
L. Xu, F. Ding, Y. Gu et al., A multi-innovation state and parameter estimation algorithm for a state space system with d-step state-delay. Signal Process. 140, 97–103 (2017)
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix 1
Assume similarity transformation matrix \( {\varvec{S}} \in {\mathbb{R}}^{(\tau + 1)n \times (\tau + 1)n} \) converts the estimated augmented matrices \( {\varvec{\hat{\bar{A}}}} \), \( {\varvec{\hat{\bar{B}}}} \), \( {\varvec{\hat{\bar{C}}}} \) to the following desired form,
where \( {\hat{\varvec{A}}}_{{\varvec{1}}}^{{\varvec{d}}} , \, {\hat{\varvec{A}}}_{{\varvec{2}}}^{{\varvec{d}}} \in {\mathbb{R}}^{n \times n} \), \( {\hat{\varvec{B}}}^{{\varvec{d}}} \in {\mathbb{R}}^{n \times m} \) and \( {\hat{\varvec{C}}}^{{\varvec{d}}} \in {\mathbb{R}}^{p \times n} \). Therefore, we have,
Let us represent matrix \( {\varvec{S}} \) as
where \( {\varvec{S}}_{{\varvec{1}}} , \, {\varvec{S}}_{{\varvec{2}}} , \, \ldots { , }{\varvec{S}}_{{{\varvec{\tau + 1}}}} \in {\mathbb{R}}^{n \times (\tau + 1)n} \). Then, Eq. (50) implies
Since \( {\hat{\varvec{A}}}_{{\varvec{1}}}^{{\varvec{d}}} \) and \( {\hat{\varvec{A}}}_{{\varvec{2}}}^{{\varvec{d}}} \) are full matrices, Eq. (55) only causes the condition
Now, we investigate Eq. (51). It is obvious from Eq. (51) that
Equations (57) and (54) imply that
It can be concluded from Eq. (58) that the row vectors of matrix \( {\varvec{S}}_{{\varvec{1}}} \) are in the space spanned through the row vectors of \( \left( {{\text{null}}\{ [\begin{array}{*{20}c} {{\varvec{\hat{\bar{B}}}}} & {{\varvec{\hat{\bar{A}}\hat{\bar{B}}}}} & \ldots & {{\varvec{\hat{\bar{A}}}}^{\tau - 1} {\varvec{\hat{\bar{B}}}}} \\ \end{array} ]^{T} \} } \right)^{T} \), where \( {\text{null}}\{ .\} \) is the null space of a matrix. Let us define,
in which, \( q = (\tau + 1)n{\text{ - columnrank}}\{ [\begin{array}{*{20}c} {{\varvec{\hat{\bar{B}}}}} & {{\varvec{\hat{\bar{A}}\hat{\bar{B}}}}} & \ldots & {{\varvec{\hat{\bar{A}}}}^{\tau - 1} {\varvec{\hat{\bar{B}}}}} \\ \end{array} ]\} \) and \( {\varvec{F}}_{{\varvec{1}}} , { }{\varvec{F}}_{{\varvec{2}}} , \, \ldots { , }{\varvec{F}}_{{\varvec{q}}} \in {\mathbb{R}}^{1 \times (\tau + 1)n} \) are the row vectors of matrix \( {\varvec{F}} \). Then, \( {\varvec{S}}_{{\varvec{1}}} \) must be as,
where \( {\varvec{Z}} \in {\mathbb{R}}^{n \times q} \).
Finally, investigating Eq. (52) implies,
Therefore, according to Eqs. (53), (54), (56), (60) and (61), the similarity transformation matrix \( {\varvec{S}} \) which transforms the identified augmented matrices to the desired form (49) can be considered as,
where \( {\varvec{Z}} \in {\mathbb{R}}^{n \times q} \) is a matrix such that the conditions
be satisfied.
Now, we find a solution \( {\varvec{Z}} \in {\mathbb{R}}^{n \times q} \) for the special case \( \tau = 1 \). In this case, the similarity transformation is as
where \( {\varvec{Z}} \in {\mathbb{R}}^{n \times q} \) is a matrix such that \( {\varvec{\hat{\bar{C}}S}}^{ - 1} (:,1:n) = {\varvec{0}}_{p \times n} \). Therefore, the row vectors of matrix \( {\varvec{\hat{\bar{C}}}} \) must be in the space spanned through the row vectors of \( \left( {{\text{null}}\{ ({\varvec{S}}^{ - 1} (:,1:n))^{T} \} } \right)^{T} \). From \( {\varvec{SS}}^{ - 1} = {\varvec{I}} \), we have \( {\varvec{ZF\hat{\bar{A}}S}}^{ - 1} (:,1:n) = {\varvec{0}}_{n \times n} \). Since \( {\text{rank}}({\varvec{ZF\hat{\bar{A}}}}) = n \), \( {\varvec{ZF\hat{\bar{A}}}} \) can be considered as \( \left( {{\text{null}}\{ ({\varvec{S}}^{ - 1} (:,1:n))^{T} \} } \right)^{T} \). In summary,
where \( {\varvec{\beta}} \in {\mathbb{R}}^{p \times n} \).
Assume
where \( {\varvec{\beta}}_{{\varvec{1}}} \in {\mathbb{R}}^{l \times n} \) and \( {\varvec{\hat{\bar{C}}}}_{{\varvec{s}}} \in {\mathbb{R}}^{l \times 2n} \) contains the independent rows of \( {\varvec{\hat{\bar{C}}}} \), then Eq. (62) is satisfied. Without loss of generality, we consider \( {\varvec{\beta}}_{{\varvec{1}}} = [\begin{array}{*{20}c} {{\varvec{I}}_{{\varvec{l}}} } & {{\varvec{0}}_{{{\varvec{l \times (n - l)}}}} } \\ \end{array} ] \), then \( {\varvec{Z}}(1:l,:) = {\varvec{\hat{\bar{C}}}}_{{\varvec{s}}} /({\varvec{F\hat{\bar{A}}}}) \). There is no constraint on \( {\varvec{Z}}(l + 1:n,:) \in {\mathbb{R}}^{(n - l) \times q} \) unless it makes \( {\varvec{Z}} \) and \( {\varvec{S}} \) full rank.
Appendix 2
Let the characteristic polynomial of \( {\bar{\varvec{A}}} \) be given as
and
Then, using the Cayley–Hamilton theorem, we derive
Premultiplying Eq. (7) with \( {{\varvec{\Gamma}}_{{{\varvec{CP}}}}^{ \bot }}^{T} \) implies
which can be rewritten as
Define
where \( \hat{\alpha }_{i} , \, i = 1, \ldots ,\hat{n}_{1} \) are the coefficients of the characteristic polynomial of \( {\varvec{\hat{\bar{A}}}} \), and \( \mathop{{\hat{\varvec{\Gamma }}_{{\varvec{CP}}}^{ \bot }}}\nolimits^{T} \) and \( {\hat{\varvec{\psi }}}_{{\varvec{k}}} \) are constructed similar to \( {{\varvec{\Gamma}}_{{{\varvec{CP}}}}^{ \bot }}^{T} \) and \( {\varvec{\psi}}_{{\varvec{k}}} \), respectively, but with the coefficients \( \hat{\alpha }_{i} \) and the estimated matrices \( {\varvec{\hat{\bar{B}}}} \), \( {\varvec{\hat{\bar{C}}}} \) and \( {\hat{\varvec{D}}}^{{\varvec{d}}} \).
Equations (71) and (72) concludes,
Appendix 3
Rights and permissions
About this article
Cite this article
Ahmadipour, M., Seydnejad, S.R. & Barkhordari-Yazdi, M. Subspace-Based Deterministic Identification of MIMO Linear State-Delayed Systems. Circuits Syst Signal Process 39, 4067–4093 (2020). https://doi.org/10.1007/s00034-020-01355-4
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00034-020-01355-4