Abstract
In this paper, a secure chaotic scheme for communications in noisy public channel is proposed. This scheme is based on the concept of carrier encryption in addition to the typical data encryption techniques. At the transmitter end, the unified chaotic system with adaptive parameter and a hyperchaotic Rössler system with uncertain parameters are coupled, constrained and used as a new hyperchaotic system which generates waveforms that are different from those of any known chaotic oscillator. After modulating one of the outputs of the system with the encrypted data signal, the outputs of the system are encrypted using a set of pre-defined encryption rules and transmitted to the receiving end through a noisy public communications channel. At the receiving end, the received outputs are decrypted and the transmitted data are retrieved by reconstructing the constrained hyperchaotic signals using the novel discrete-time iterative decomposed uncertain constrained extended Kalman filter (IDUCEKF). The proposed state estimator, besides being used to handle the estimation problem of uncertain constrained nonlinear dynamical systems, reduces the required processing time and gives good numerical performance. Simulation results are firstly presented to illustrate the applicability of the IDUCEKF in synchronizing the states of the constrained hyperchaotic system. Then, the proposed secure communication scheme is applied to transmit images, and the quality of the transmission process is assessed. The obtained results show the effectiveness of the proposed approach.
Similar content being viewed by others
References
Satish, K., Jayakar, T., Tobin, C., Kadhavi, K., Murali, K.: Chaos based spread spectrum image steganography. IEEE Trans. Consum. Electron. 50, 587–590 (2004)
Fallahi, K., Raoufi, R., Khoshhin, H.: An application of Chen system for secure chaotic communication based on extended Kalman filter and multi-shift cipher algorithm. Nonlinear Sci. Numer. Simul. 13, 763–781 (2008)
Heidari Bateni, G., Mc Gillem, C.D.: A chaotic direct sequence spread spectrum communication system. IEEE Trans. Commun. 42, 1524–1527 (1994)
Kolumban, G., Kennedy, M.P., Chua, L.O.: The role of synchronization in digital communications using chaos—part I: fundamentals of digital communications. IEEE Trans. Circuits Syst. I 44, 927–936 (1998)
Kolumban, G., Kennedy, M.P., Chua, L.O.: The role of synchronization in digital communications using chaos—part II: chaotic modulation and chaotic synchronization. IEEE Trans. Circuits Syst. I 45, 1129–1140 (1998)
Kolumban, G., Kennedy, M.P., Chua, L.O.: The role of synchronization in digital communications using chaos—part III: performance bounds for correlation receivers. IEEE Trans. Circuits Syst. I 47, 1673–1683 (2000)
Pecora, L.M., Carroll, T.L., Johnson, G.A., Mar, D.J.: Fundamentals of synchronization in chaotic systems, concepts, and applications. Chaos 7, 520–542 (1997)
Murali, K.: Digital signal transmission with cascaded heterogeneous chaotic systems. Phys. Rev. E 63, 016217–23 (2001)
Calitoiu, D., Oommen, B.J., Nussbaum, D.: Desynchronizing a chaotic pattern recognition neural network to model inaccurate perception. IEEE Trans. Syst. Man Cybern. B 37, 692–704 (2007)
Ruan, H., Zhai, T., Yaz, E.E.: A chaotic secure chaotic communication scheme with extended Kalman filter based parameter estimation. Proc. IEEE Conf. Control Appl. 1, 404–408 (2003)
Leung, H., Zhu, Z., Ding, Z.: An aperiodic phenomenon of the extended Kalman filter in filtering noisy chaotic signals. IEEE Trans. Signal Process. 48, 1807–1810 (2000)
Li, S.Y., Ge, Z.M.: Fuzzy modelling and synchronization of two totally different chaotic systems via novel fuzzy mode. IEEE Trans. Syst. Man Cybern. B 41, 1015–1026 (2011)
Zhang, H., Ma, T., Huang, G.-B., Wang, Z.: Robust global exponential synchronization of uncertain chaotic delayed neural network via dual-stage impulsive control. IEEE Trans. Syst. Man Cybern. B 40, 831–844 (2010)
Ahn, C.K.: Takagi–Sugeno fuzzy receding horizon H\(\infty \) chaotic synchronization and its application to the Lorenz system. Nonlinear Anal. Hybrid Syst. 9, 1–8 (2013)
Shi, X., Wang, Z.: Robust chaos synchronization of four-dimensional energy resource system via adaptive feedback control. Nonlinear Dyn. 60, 631–637 (2010)
Lu, J.A., Wu, X.Q., Lu, J.H.: Synchronization of a unified chaotic system and the application in secure communication. Phys. Lett. A 305, 365–370 (2002)
Chen, H.H., Sheu, G.J., Lin, Y.L., Chen, C.S.: Chaos synchronization between two different chaotic systems via nonlinear feedback control. Nonlinear Anal. Theory Methods Appl. 70, 4393–4401 (2009)
Wang, X., Song, J.: Synchronization of the unified chaotic system. Nonlinear Anal. Theory Methods Appl. 69, 3409–3416 (2008)
Lin, J.S., Yan, J.J.: Adaptive synchronization for two identical generalized Lorenz chaotic systems via a single controller. Nonlinear Anal. Real World Appl. 10, 1151–1159 (2009)
Njah, A.N.: Tracking control and synchronization of the new hyperchaotic Liu system via backstepping techniques. Nonlinear Dyn. 61(1–2), 1–9 (2010)
Xiang-Jun, W., Jing-Sen, L., Guan-Rong, C.: Chaos synchronization of Rikitake chaotic attractor using the passive control technique. Nonlinear Dyn. 53(1–2), 45–53 (2008)
Ablay, G.: Sliding mode control of uncertain unified chaotic systems. Nonlinear Anal. Hybrid Syst. 3, 531–535 (2009)
Smaoui, N., Karouma, A., Zribi, M.: Secure communication based on the synchronization of the hyperchaotic Chen and the unified chaotic systems. Commun. Nonlinear Sci. Numer. Simul. 16, 3279–3293 (2011)
Mata-Machuca, J.L., MartÃnez-Guerra, R., Aguilar-López, R., Aguilar-Ibañez, C.: A chaotic system in synchronization and secure communications. Commun. Nonlinear. Sci. Numer. Simul. 17, 1706–1713 (2012)
Wang, H., Han, Z.Z., Zhang, W., Xie, Q.Y.: Synchronization of unified chaotic systems with uncertain parameters based on the CLF. Nonlinear Anal. Real World Appl. 10, 715–722 (2009)
Wang, H., Han, Z.Z., Xie, Q.Y., Zhang, W.: Finite-time synchronization of uncertain chaotic systems based on CLF. Nonlinear Anal. Real World Appl. 10, 2842–2849 (2009)
Nosrati, K., Azemi, A., Pariz, N., Shokouhi-R, A.: Chaotic synchronization of Lorenz system using Unscented Kalman Filter. Proc. Chin. Control Decis. Conf. 1, 848–853 (2011)
Chen, S.H., Yang, Q., Wang, C.P.: Impulsive control and synchronization of unified chaotic system. Chaos Solitons Fract. 20, 153–160 (2004)
Yang, X., Yang, Z., Nie, X.: Exponential synchronization of discontinuous chaotic systems via delayed impulsive control and its application to secure communication. Commun. Nonlinear Sci. Numer. Simul. 19, 1529–1543 (2014)
Zhu, Z., Leung, H.: Adaptive blind equalization for chaotic communication systems using extended Kalman filter. IEEE Trans. Circuits Syst. I(48), 979–987 (2001)
Zhu, F.: Observer based synchronization of uncertain chaotic system and its application to secure communications. Chaos Solitons Fract. 40, 2384–2391 (2009)
Yang, J., Zhu, F.: Synchronization for chaotic systems and chaos-based secure communications via both reduced-order and step-by-step sliding mode observers. Commun. Nonlinear Sci. Numer. Simul. 18, 926–937 (2013)
Chen, M., Zhou, D., Shang, Y.: A sliding mode observer based secure communication scheme. Chaos Solitons Fract. 25, 573–578 (2005)
Short, K.M.: Unmasking a modulated chaotic communication scheme. Int. J. Bifurc. Chaos 6, 367–375 (1996)
Perez, G., Cerdeira, H.A.: Extracting messages masked by chaos. Phys. Rev. Lett. 74, 1970–1973 (1995)
Tlelo-Cuautle, E., Carbajal-Gomez, V.H., Obeso-Rodelo, P.J., Rangel-Magdaleno, J.J., Núñez-Pérez, J.C.: FPGA realization of a chaotic communication system applied to image processing. Nonlinear Dyn. (2015). doi:10.1007/s11071-015-2284-x
Tlelo-Cuautle, E., Rangel-Magdaleno, J.J., Pano-Azucena, A.D., Obeso-Rodelo, P.J., Nunez-Perez, J.C.: FPGA realization of multi-scroll chaotic oscillators. Commun. Nonlinear Sci. Numer. Simul. 27(1), 66–80 (2015)
GarcÃa-MartÃnez, M., Campos-Cantón, E.: Pseudo-random bit generator based on multi-modal maps. Nonlinear Dyn. (2015). doi:10.1007/s11071-015-2303-y
De la Fraga, L.G., Tlelo-Cuautle, E.: Optimizing the maximum Lyapunov exponent and phase space portraits in multi-scroll chaotic oscillators. Nonlinear Dyn. 76(2), 1503–1515 (2014)
Hassan, M.F.: Optimal Kalman filter for large scale systems using the partitioning approach. IEEE Trans. Syst. Man Cyber. 6, 714–720 (1976)
Hassan, M.F., Salut, G., Singh, M.G., Titli, A.: A decentralized computational algorithm for the global Kalman filter. IEEE Trans. Automat. Control 23, 262–268 (1978)
Mahmoud, M.S., Hassan, M.F., Darwish, M.G.: Large-Scale Control Systems: Theories and Techniques. Marcel Dekker Inc., New York (1985)
Hassan, M.F.: A Decomposed Estimator for Constrained Uncertain Stochastic Nonlinear Systems (submitted for publication) (2015)
Zhou, Q., Shi, P., Liu, H., Xu, S.: Neural-network-based decentralized adaptive output-feedback control for large-scale stochastic nonlinear systems. IEEE Trans. Syst. Man Cyber. B 42, 1608–1619 (2012)
Liu, S., Zhang, J., Jiang, Z.: Decentralized adaptive output-feedback stabilization for large-scale stochastic nonlinear systems. Automatica 43, 238–251 (2007)
Mahmoud, M.S.: Decentralized Control and Filtering in Interconnected Dynamical Systems. CRC Press, Boca Raton (2011)
Wang, X.Y., Wu, X.J.: Tracking control and synchronization of four-dimensional hyperchaotic Rössler system. Chaos 16, 033121 (2006)
Meditch, J.S.: Stochastic Optimal Linear Estimation and Control. McGraw-Hill, New York (1965)
Reif, K., Gunther, S., Yaz, E., Unbehauen, R.: Stochastic stability of the discrete-time extended Kalman filter. IEEE Trans. Automat. Control 44, 714–728 (1999)
Hassan, M.F., Alrifai, M.T., Soliman, H.M., Kourah, M.A.: Observer-based controller for constrained uncertain stochastic nonlinear discrete-time systems. Int. J. Robust Nonlinear Control (2015). doi:10.1002/rnc.3396
Luenberger, D.: Optimization by Vector Space Methods. Wiley, New York (1969)
Hassan, M.F.: Iterated constrained state estimator for nonlinear discrete-time systems with uncertain parameters. Int. J. Innov. Comput. Inf. Control 8, 6141–6160 (2012)
Sandri, M.: Numerical calculation of Lyapunov exponents. Math. J. 6, 78–84 (1996)
Eckmann, J.P., Ruelle, D.: Ergodic theory of chaos and strange attractors. Rev. Mod. Phys. 57, 617–656 (1985)
Holte, J.M.: Discrete Gronwall lemma and applications, MAA-NCS Meeting at the University of North Dakota 24 (2009). http://homepages.gac.edu/~holte/publications/GronwallLemma.pdf
Acknowledgments
This research was supported by Kuwait University under research Grant No. EE 02/14.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix 1
Proof of Lemma 4.1
For the ith subsystem, by linearizing the system dynamics around \(\hat{\varvec{x}}({\varvec{k}}| {\varvec{k}})\) and \(\bar{\varvec{b}}\) while neglecting higher-order terms, we get:
where \({\hat{\varvec{F}}}_{\varvec{i}}^{\prime } \), \({\hat{\varvec{E}}}_{\varvec{i}}^{\prime } \) are as defined by (18).
Substituting (83) into the ith subsystem model (14), we get:
The conditional expectation of (84) given the measurements up to the kth instant of time is given by:
where \({\varvec{Y}}^{{\varvec{k}}}=[{\varvec{y}}^{{\varvec{T}}}(1) {\varvec{y}}^{{\varvec{T}}}(2) \cdots {\varvec{y}}^{{\varvec{T}}}({\varvec{k}})]\)
Since \(\hat{\varvec{x}}({\varvec{k}}| {\varvec{k}})\cong {\varvec{E}}\left\{ {{{\varvec{x}}({\varvec{k}})} |{\varvec{Y}}^{{\varvec{k}}}} \right\} \), the variables \({\varvec{Y}}^{{\varvec{k}}}\), \({\varvec{w}}_{\varvec{i}} ({\varvec{k}})\) and \({\varvec{b}}({\varvec{k}})\) are independent, \({\varvec{E}}\left\{ {{\varvec{w}}_{\varvec{i}} ({\varvec{k}})} \right\} =0\) and \({\varvec{E}}\left\{ {{\varvec{b}}({\varvec{k}})} \right\} =\bar{\varvec{b}}\), then the predicted estimate \(\hat{\varvec{x}}_{\varvec{i}} ({\varvec{k}}+1| {\varvec{k}})\) takes the form:
let \({\tilde{\varvec{x}}}({\varvec{k}}| {\varvec{k}})={\varvec{x}}({\varvec{k}})-\hat{\varvec{x}}({\varvec{k}}| {\varvec{k}})\), \({\tilde{\varvec{x}}}_{\varvec{i}} ({\varvec{k}}+1| {\varvec{k}})={\varvec{x}}_{\varvec{i}} ({\varvec{k}})-\hat{\varvec{x}}_{\varvec{i}} ({\varvec{k}}+1| {\varvec{k}})\), and \(\tilde{{\varvec{b}}}({\varvec{k}})={\varvec{b}}({\varvec{k}})-\bar{\varvec{b}}\), then using (84) and (86), we get:
The covariance matrix of the prediction error of the ith subsystem is defined by:
Substituting from (87) into (88) and with the use of model properties (13) (i.e., \({\varvec{w}}({\varvec{k}})\), and \(\tilde{\varvec{b}}({\varvec{k}})\) are independent, \({\varvec{x}}_{\varvec{i}} ({\varvec{k}})\) is independent of \({\varvec{w}}({\varvec{k}})\) and \({\varvec{b}}({\varvec{k}})\), and \({\varvec{E}}\left\{ {\tilde{\varvec{b}}({\varvec{k}})} \right\} ={\varvec{E}}\left\{ {{\varvec{b}}({\varvec{k}})-\bar{\varvec{b}}} \right\} =0)\), one gets after simple mathematical manipulations:
The cross-covariance matrix of the prediction error of the ith and the jth subsystems is defined by:
With the assumptions of the model properties and the independency of the noise subvectors \({\varvec{w}}_{\varvec{i}} ({\varvec{k}})\) and \({\varvec{w}}_{\varvec{j}} ({\varvec{k}})\), it is easy to show that
Equation (86) gives the predicted estimate of \({\varvec{x}}_{\varvec{i}} ({\varvec{k}}+1)\) denoted by \(\hat{\varvec{x}}_{\varvec{i}} ({\varvec{k}}+1| {\varvec{k}})\) for \(i\in \{1,2,{\ldots },N\}\), Eq. (89) gives the covariance matrices of the prediction estimation error of \({\varvec{x}}_{\varvec{i}} ({\varvec{k}})\), whereas Eq. (91) gives the cross-covariance matrices of the estimation errors of \({\varvec{x}}_{\varvec{i}} ({\varvec{k}})\) and \({\varvec{x}}_{\varvec{j}} ({\varvec{k}}) \quad \forall \; i,j \in \left\{ {i+1, 2, \ldots ,N} \right\} \).
Now, assume that we received the measurement vector \({\varvec{y}}({\varvec{k}}+1)\) at the \((k+1)\)th sampling instant. By decomposing \({\varvec{y}}({\varvec{k}}+1)\in {\varvec{R}}^{{\varvec{m}}}\) into M subvectors \({\varvec{y}}_{\varvec{r}} ({\varvec{k}}+1)\in {\varvec{R}}^{{\varvec{m}}_{\varvec{r}}}\) where \(r\in \{1,2,\ldots , M\}\) and \(\sum _{r=1}^M {m_r =m} \), then the predicted estimates of the state subvectors \(\hat{\varvec{x}}_{\varvec{i}} ({\varvec{k}}+1| {\varvec{k}})\) for \(i\in \{1,2,\ldots , N\}\) will be updated successively using the measurement subvectors \({\varvec{y}}_{\varvec{r}} ({\varvec{k}}+1)\) for \(r\in \{1,2,\ldots , M\}\). Therefore, we have
where
Using the multiple projection approach [41, 42], we have:
where for \(r=1\) we have:
and for \(r \in \{2, 3, \ldots ,M\}\), we have:
Using the first submeasurement vector \({\varvec{y}}_1({\varvec{k}}+1)\), the approximated filtered estimate \(\hat{\varvec{x}}^{1}_{\varvec{i}} ({\varvec{k}}+1| {{\varvec{k}}+1})\) for \(i\in \{1,2,\ldots , N\}\) is such that:
where
Since \({\varvec{x}}({\varvec{k}}+1)=\hat{\varvec{x}}({\varvec{k}}+1| {\varvec{k}})+{\tilde{\varvec{x}}}({\varvec{k}}+1| {\varvec{k}})\) and \({\varvec{c}}({\varvec{k}}+1)=\bar{\varvec{c}}+\tilde{{\varvec{c}}}({\varvec{k}}+1)\), and assuming that \({\varvec{x}}({\varvec{k}}+1)\) is close to \(\hat{\varvec{x}}({\varvec{k}}+1| {\varvec{k}})\) and \({\varvec{c}}({\varvec{k}}+1)\) is close to \(\bar{\varvec{c}}\) such that \({\varvec{h}}_1 ({\varvec{x}}({\varvec{k}}+1),{\varvec{c}}({\varvec{k}}+1))\) can be replaced by its first-order approximation, i.e., we assume \({\varvec{h}}_1 ({\varvec{x}}({\varvec{k}}+1),{\varvec{c}}({\varvec{k}}+1))\) is affine on a neighborhood of \({\varvec{x}}({\varvec{k}}+1),\hat{\varvec{x}}({\varvec{k}}+1| {\varvec{k}})\) and \({\varvec{c}}({\varvec{k}}+1), \bar{\varvec{c}}\). Then, we have:
where \(\hat{\varvec{H}}_1 ({\varvec{k}}+1)=\hat{\varvec{H}}^{0}_1 ({\varvec{k}}+1), \hat{\varvec{D}}_1 ({\varvec{k}}+1)=\hat{\varvec{D}}_1^0 ({\varvec{k}}+1)\) are defined by (26) for \(r=1\).
Substituting from (100) into (98) while using (95), we get:
The cross-covariance matrix \({\varvec{P}}_{{\varvec{x}}_{\varvec{i}} {\tilde{\varvec{y}}}_1} ({\varvec{k}}+1| {\varvec{k}})\) is defined by:
From the Gaussian conditional estimator [48], since \({\varvec{E}}\left\{ {\hat{\varvec{x}}_{\varvec{i}} ({\varvec{k}}+1| {\varvec{k}}) {\tilde{\varvec{y}}}^{{\varvec{T}}}_1 ({\varvec{k}}+1| {\varvec{k}})} \right\} \cong 0\), then we have:
Since \({\varvec{E}}\left\{ {\tilde{{\varvec{c}}}({\varvec{k}}+1)} \right\} =0\), \({\varvec{E}}\left\{ {{\varvec{v}}_1 ({\varvec{k}}+1)} \right\} =0\) and \({\tilde{\varvec{x}}}({\varvec{k}}+1| {\varvec{k}})\) is independent of both \(\tilde{{\varvec{c}}}({\varvec{k}}+1)\) and \({\varvec{v}}_1 ({\varvec{k}}+1)\), we get:
Using (101) and recalling that \({\tilde{\varvec{x}}}({\varvec{k}}+1| {\varvec{k}})\), \(\tilde{{\varvec{c}}}({\varvec{k}}+1)\), and \({\varvec{v}}_1 ({\varvec{k}}+1)\) are mutually independent, we get:
where \({\varvec{U}}({\varvec{k}}+1)\) is the covariance matrix of the output model parameter \({\varvec{c}}({\varvec{k}}+1)\) and \({\varvec{R}}_{11} ({\varvec{k}}+1)\) is the covariance matrix of the output noise subvector \({\varvec{v}}_1 ({\varvec{k}}+1)\).
Therefore, substituting from (102) and (103) into (99) while using (22), the gain matrix \({\varvec{K}}_{1{\varvec{i}}}^1 ({\varvec{k}}+1)\) is given by:
The covariance matrix of the estimation error of the filtered estimate \(\hat{\varvec{x}}^{1}_{\varvec{i}} ({\varvec{k}}+1| {{\varvec{k}}+1})\) denoted by \({\varvec{P}}_{{\tilde{\varvec{x}}}_{\varvec{i}} {\tilde{\varvec{x}}}_{\varvec{i}} }^1 ({\varvec{k}}+1| {{\varvec{k}}+1})\) is given by:
where the error \({\tilde{\varvec{x}}}^{1}_i ({\varvec{k}}+1| {{\varvec{k}}+1})\) is defined as:
Substituting from (106) into (105) while using (99) and after simple mathematical manipulation, it is easy to get:
The cross-covariance matrix of the error of the filtered estimates \(\hat{\varvec{x}}^{1}_{\varvec{i}} ({\varvec{k}}\!+\!1| {{\varvec{k}}+1})\) and \(\hat{\varvec{x}}^{1}_{\varvec{j}} ({\varvec{k}}+1| {{\varvec{k}}+1})\) denoted by \({\varvec{P}}_{{\tilde{\varvec{x}}}_{\varvec{i}} {\tilde{\varvec{x}}}_{\varvec{j}} }^1 ({\varvec{k}}+1| {{\varvec{k}}+1})\) is given by:
where \({\tilde{\varvec{x}}}^{1}_i ({\varvec{k}}+1| {{\varvec{k}}+1})\) is defined by (106) and \({\tilde{\varvec{x}}}^{1}_{\varvec{j}} ({\varvec{k}}+1| {{\varvec{k}}+1})\) is given by:
Again, using (106) and (108), it is easy to show that the cross-covariance matrices for \(j\in \{i+1,\ldots ,N\} \,\, {\varvec{P}}_{{\tilde{\varvec{x}}}_{\varvec{i}} {\tilde{\varvec{x}}}_{\varvec{j}}}^1 ({\varvec{k}}+1| {{\varvec{k}}+1})\) are such that:
Following the same procedure, it can be shown that the filtered estimate \(\hat{\varvec{x}}^{{\varvec{r}}}_{\varvec{i}} ({\varvec{k}}+1| {{\varvec{k}}+1})\) corresponding to the measurements \({\varvec{Y}}^{{\varvec{k}}}, {\varvec{y}}_1 ({\varvec{k}}+1), {\varvec{y}}_2 ({\varvec{k}}+1),\ldots , {\varvec{y}}_{\varvec{r}} ({\varvec{k}}+1)\), its covariance matrix \({\varvec{P}}_{{\tilde{\varvec{x}}}_{\varvec{i}} {\tilde{\varvec{x}}}_{\varvec{i}} }^{\varvec{r}} ({\varvec{k}}+1| {{\varvec{k}}+1})\), and the cross-covariance matrices \({\varvec{P}}_{{\tilde{\varvec{x}}}_{\varvec{i}} {\tilde{\varvec{x}}}_{\varvec{j}} }^{\varvec{r}} ({\varvec{k}}+1| {{\varvec{k}}+1})\) are given by:
where
\(\square \)
Appendix 2
Proof of Theorem 6.1
-
1a.
Equation (67) can be written in the following alternative form:
$$\begin{aligned} {\varvec{Z}}({\varvec{k}}+1)={\varvec{\varPhi }}_{{\varvec{ss}}}^{{\varvec{k}}+1} {\varvec{Z}}(0) \end{aligned}$$Then, the norm of the expected value of the above equation is given by:
$$\begin{aligned} \bar{{\bar{{{\varvec{Z}}}}}}({\varvec{k}}+1)\le \Vert {{\varvec{\varPhi }} _{{\varvec{ss}}}^{{\varvec{k}}+1} } \Vert \bar{{\bar{{{\varvec{Z}}}}}}(0) \end{aligned}$$(115)Since for all the solution of (67) \(\bar{{\bar{{{\varvec{Z}}}}}}({\varvec{k}}+1)\) remains bounded as \({{k}}\rightarrow \infty \), then there exists a constant \({{c}}_1\) such that for \({{k}}+1\ge 0\):
$$\begin{aligned} \Vert {{\varvec{\varPhi }} _{{\varvec{ss}}}^{{\varvec{k}}+1} } \Vert \le {\varvec{c}}_1 \end{aligned}$$(116)Equation (68) can be written in the form:
$$\begin{aligned} {\varvec{Z}}({\varvec{k}}\!+\!\!1)\!=\!{\varvec{\varPhi }}_{{\varvec{ss}}}^{{\varvec{k}}+1} {\varvec{Z}}(0)\!+\!\!\sum \limits _{{\varvec{j}}=1}^{{\varvec{k}}-1} {{\varvec{\varPhi }}_{{\varvec{ss}}}^{{\varvec{k}}-{\varvec{j}}} {\varvec{\varPsi }} ({\varvec{j}}\!+\!1,{\varvec{j}}){\varvec{Z}}({\varvec{j}})} \end{aligned}$$Taking the norm of the expected value of this equation while using (70), we have:
$$\begin{aligned}&\bar{{\bar{{{\varvec{Z}}}}}}_ ({\varvec{k}}+1)\le \Vert {{\varvec{\varPhi }}_{{\varvec{ss}}}^{{\varvec{k}}+1} } \Vert \bar{{\bar{{{\varvec{Z}}}}}}(0)\nonumber \\&\quad +\,\sum \limits _{{\varvec{j}}=0}^{{\varvec{k}}-1} {\Vert {{\varvec{\varPhi }}_{{\varvec{ss}}}^{{\varvec{k}}-{\varvec{j}}} } \Vert } \sqrt{{\varvec{\varTheta }}^{{\varvec{T}}}{\varvec{\varTheta }}}\le {\varvec{c}}_1 \bar{{\bar{{{\varvec{Z}}}}}}(0)\nonumber \\&\quad +\,\sum \limits _{{\varvec{j}}=0}^{{\varvec{k}}-1} {{\varvec{c}}_1 } {\varvec{\lambda }} ({\varvec{j}})\bar{{\bar{{{\varvec{Z}}}}}}({\varvec{j}}) \end{aligned}$$(117)Let \(g(j)=c_1 \lambda (j)\),
$$\begin{aligned} \therefore \,\,\bar{{\bar{{{\varvec{Z}}}}}}({\varvec{k}}+1)\le {\varvec{c}}_1 \bar{{\bar{{{\varvec{Z}}}}}}(0)+\sum \limits _{{\varvec{j}}=0}^{{\varvec{k}}-1} {{\varvec{g}}({\varvec{j}})} \bar{{\bar{{{\varvec{Z}}}}}}({\varvec{j}}) \end{aligned}$$From the special Gronwall lemma [55], we have:
$$\begin{aligned} \bar{{\bar{{{\varvec{Z}}}}}}({\varvec{k}}+1)\le & {} {\varvec{c}}_1 \bar{{\bar{{{\varvec{Z}}}}}}(0)\prod _{{\varvec{j}}=0}^{{\varvec{k}}-1} (1+{\varvec{g}}({\varvec{j}}))\nonumber \\\le & {} {\varvec{c}}_1 \bar{{\bar{{{\varvec{Z}}}}}}(0)\exp \left( \sum \limits _{{\varvec{j}}=0}^{{\varvec{k}}-1} {{\varvec{g}}({\varvec{j}})} \right) \end{aligned}$$(118)By the hypothesis (69), there exists a constant \(c_2\) such that \(\sum \limits _{k=0}^\infty {\lambda (k)} \le c_2 \). Therefore, we have:
$$\begin{aligned} \bar{{\bar{{{\varvec{Z}}}}}}({\varvec{k}}+1)\le {\varvec{c}}_1 \bar{{\bar{{{\varvec{Z}}}}}}(0)\exp ({\varvec{c}}_1 {\varvec{c}}_2 ) \end{aligned}$$(119)which shows that the expected value of the norm of the homogeneous system is bounded. Since \({\varvec{E}}\{{\varvec{w}}({\varvec{k}})\}={\varvec{E}}\{\tilde{\varvec{b}}({\varvec{k}})\}={\varvec{E}}\{ \tilde{{\varvec{c}}}({\varvec{k}}+1)\}={\varvec{E}}\{{\varvec{v}}({\varvec{k}}+1)\}=0\), then \({\varvec{E}}\{{\varvec{\mu }} ({\varvec{k}})\}=0\) in (64) or (66). Therefore, the norm of the expected value of (66) as given by (117) is bounded according to (119).
-
1b.
Since the expected value of the constrained equilibrium \({\varvec{x}}_{{\varvec{ss}}} \)is detectable, and the pair \(\left( {\hat{{\varvec{F}}}^{\prime }\!({\varvec{k}}), \hat{\varvec{H}}({\varvec{k}}+1)} \right) \) is detectable for the entire domain \({\varvec{\varOmega }}\) except for a finite number of points, then the matrix \({\varvec{\varPhi }}_{{\varvec{ss}}}\) is stable. Therefore, the origin is the zero equilibrium of the expected value of the estimation error. As a result, there exist \(0\le {\varvec{\delta }}({\varvec{k}}+1) <1\) such that \(\Vert {{\varvec{\varPhi }} _{{\varvec{ss}}}^{{\varvec{k}}+1} } \Vert <[{\varvec{\delta }} ({\varvec{k}}+1)]^{{\varvec{k}}+1}\) for \({{k}}+1\ge {{K}}\).
In this case, i.e., \(k+1\ge K\), and from the Gronwall lemma while using (69), the inequality (117) takes the form:
$$\begin{aligned} \bar{{\bar{{{\varvec{Z}}}}}}({\varvec{k}}+1)\le & {} \left\| {{\varvec{\varPhi }}_{{\varvec{ss}}}^{{\varvec{k}}+1} } \right\| \bar{{\bar{{{\varvec{Z}}}}}}(0)+\sum \limits _{{\varvec{j}}=0}^{{\varvec{k}}-1} {\left\| {{\varvec{\varPhi }}_{ss}^{k-j} } \right\| } \sqrt{{\varvec{\varTheta }} ^{{\varvec{T}}}{\varvec{\varTheta }} }\nonumber \\\le & {} {\varvec{c}}_1 \bar{{\bar{{{\varvec{Z}}}}}}(0)+\sum \limits _{{\varvec{j}}=0}^{{\varvec{k}}-1} {{\varvec{c}}_1 } {\varvec{\lambda }} ({\varvec{j}})\bar{{\bar{{{\varvec{Z}}}}}}({\varvec{j}})\nonumber \\ \bar{{\bar{{{\varvec{Z}}}}}}({\varvec{k}}+1)\le & {} \left\| {{\varvec{\varPhi }}_{{\varvec{ss}}}^{{\varvec{k}}+1} } \right\| \bar{{\bar{{{\varvec{Z}}}}}}(0)+\sum \limits _{{\varvec{j}}=0}^{{\varvec{k}}-1} {{\varvec{g}}({\varvec{j}})} \bar{{\bar{{{\varvec{Z}}}}}}(j)\nonumber \\ \bar{{\bar{{{\varvec{Z}}}}}}({\varvec{k}}+1)\le & {} \left\| {{\varvec{\varPhi }}_{{\varvec{ss}}}^{{\varvec{k}}+1} } \right\| \bar{{\bar{{{\varvec{Z}}}}}}(0)\prod _{{\varvec{j}}=0}^{{\varvec{k}}-1} {(1+{\varvec{g}}({\varvec{j}})) } \nonumber \\\le & {} \left\| {{\varvec{\varPhi }}_{{\varvec{ss}}}^{{\varvec{k}}+1} } \right\| \bar{{\bar{{{\varvec{Z}}}}}}(0)\exp \left( \sum \limits _{{\varvec{j}}=0}^{\varvec{k}} {{\varvec{g}}({\varvec{j}})} \right) \nonumber \\ \bar{{\bar{{{\varvec{Z}}}}}}({\varvec{k}}+1)\le & {} [{\varvec{\delta }} ({\varvec{k}}+1)]^{{\varvec{k}}+1}\bar{{\bar{{{\varvec{Z}}}}}}(0)\exp ({\varvec{c}}_1 {\varvec{c}}_2 )\nonumber \\\le & {} \bar{{{\varvec{\delta }}}}^{{\varvec{k}}+1}\bar{{\bar{{{\varvec{Z}}}}}}(0)\exp ({\varvec{c}}_1 {\varvec{c}}_2 ) \end{aligned}$$(120)where \(\bar{{{\varvec{\delta }} }}=\sup \{{\varvec{\delta }} ({\varvec{j}}), {\varvec{j}}\in ({\varvec{K}},{\varvec{K}}+1,\ldots ,\infty )\}\) and \(0\le \bar{{{{\varvec{\delta }}}}}<1\)
Since \(\mathop {\lim }\limits _{{\varvec{k}}\rightarrow \infty } \bar{{{\varvec{\delta }} }}^{{\varvec{k}}+1}=0 \), then \( \mathop {\lim }\limits _{{\varvec{k}}\rightarrow \infty } \bar{{\bar{{{\varvec{Z}}}}}}({\varvec{k}}+1)=0\). Therefore, the norm of the expected value of the estimation error is locally exponentially stable.
-
2.
Equation (71) can be written as:
$$\begin{aligned} {\varvec{Z}}({\varvec{k}}+1)= & {} {\varvec{\varPhi }}_{{\varvec{ss}}}^{{\varvec{k}}+1} {\varvec{Z}}(0)+\sum \limits _{{\varvec{j}}=0}^{{\varvec{k}}-1} {\varvec{\varPhi }}_{{\varvec{ss}}}^{{\varvec{k}}-{\varvec{j}}}\nonumber \\&\times \, \left[ {{\varvec{\varPsi }} ({\varvec{j}}+1,{\varvec{j}}){\varvec{Z}}({\varvec{j}})+{\varvec{\mu }} ({\varvec{j}})} \right] \end{aligned}$$(121)Using (72), (73) and (116), the expected value of the norm of (121) is given by:
$$\begin{aligned}&\left\| {{\varvec{Z}}({\varvec{k}}+1)} \right\| _{{\varvec{av}}} \le {\varvec{c}}_1 \left\| {{\varvec{Z}}(0)} \right\| _{{\varvec{av}}}\nonumber \\&\quad +\,\sum \limits _{{\varvec{j}}=0}^{{\varvec{k}}-1} {{\varvec{c}}_1 } \left\{ {{\varvec{\lambda }} ({\varvec{j}})+{\varvec{\varepsilon }} ({\varvec{j}})} \right\} \left\| {{\varvec{Z}}({\varvec{j}})} \right\| _{{\varvec{av}}} \end{aligned}$$(122)Let \(g^{\prime } (j)=c_1 \left\{ {\lambda (j)+\varepsilon (j)} \right\} \),
$$\begin{aligned}&\therefore \,\!\left\| {{\varvec{Z}}({\varvec{k}}\!+\!1)} \right\| _{{\varvec{av}}} \!\le \!{\varvec{c}}_1 \left\| {{\varvec{Z}}(0)} \right\| _{{\varvec{av}}}\nonumber \\&\quad +\sum \limits _{{\varvec{j}}\!=\!0}^{{\varvec{k}}-1} {{\varvec{g}}^{\prime } ({\varvec{j}})} \left\| {{\varvec{Z}}({\varvec{j}})} \right\| _{{\varvec{av}}} \end{aligned}$$(123)Using Gronwall lemma, we have:
$$\begin{aligned} \left\| {{\varvec{Z}}({\varvec{k}}+1)} \right\| _{{\varvec{av}}}\le & {} {\varvec{c}}_1 \left\| {{\varvec{Z}}(0)} \right\| _{{\varvec{av}}} \prod _{{\varvec{j}}=0}^{{\varvec{k}}-1} (1+{\varvec{g}}^{\prime } ({\varvec{j}}))\nonumber \\\le & {} {\varvec{c}}_1 \left\| {{\varvec{Z}}(0)} \right\| _{{\varvec{av}}} \exp \left( \sum \limits _{{\varvec{j}}=0}^{{\varvec{k}}-1} {{\varvec{g}}^{\prime } ({\varvec{j}})} \right) \nonumber \\ \end{aligned}$$(124)By the hypothesis (72), there exists a constant \(c_3\) such that \(\sum \limits _{k=0}^\infty {\left\{ {\lambda (k)+\varepsilon (k)} \right\} } \le c_3 \). Hence, (124) is such that:
$$\begin{aligned} \left\| {{\varvec{Z}}({\varvec{k}}+1)} \right\| _{{\varvec{av}}} \le {\varvec{c}}_1 \left\| {{\varvec{Z}}(0)} \right\| _{{\varvec{av}}} \exp ({\varvec{c}}_1 {\varvec{c}}_3 ) \end{aligned}$$(125)which proves the boundedness of the expected value of the norm of the non-homogeneous system and hence the root-mean-square of the estimation error.\(\square \)
Rights and permissions
About this article
Cite this article
Hassan, M.F. Synchronization of uncertain constrained hyperchaotic systems and chaos-based secure communications via a novel decomposed nonlinear stochastic estimator. Nonlinear Dyn 83, 2183–2211 (2016). https://doi.org/10.1007/s11071-015-2474-6
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11071-015-2474-6