A unified framework for asymptotic and transient behavior of linear stochastic systems
Introduction
It is well-known that Lyapunov asymptotic stability(LAS) presents steady state behavior of systems and is very useful for practical applications. However, an asymptotic stable system may have large values of the states. Often these large values can destroy the system, which are not acceptable. For example, large transient voltage will destroy power systems [1]. To deal with these unacceptable transient values, a concept of finite-time stability(FTS) was proposed in [2], which mainly concerns the transient behavior of the system over a fixed finite-time interval and has been widely extended to deterministic linear continuous-time systems [3], [4], [5], discrete-time systems [6], stochastic systems [7], [8], Markov jump systems [9], [10], [11], [12], [13], [14] and so on. Nevertheless, up to now, there is no a unified framework to consider both transient performance and steady state performance.
On the other hand, following the development of stochastic differential equation, Itô stochastic systems have received much attention because of their practical applications, such as signal processing [15], gene networks [16], mathematical finance [17], epidemic model [18], [19]. Therefore, the related stability and control problems have been extensively studied. For example, [20] investigated mean square stability of linear stochastic systems by spectrum technique. In the sequel, literature [21] investigated input-to-state stability for stochastic nonlinear systems with state-dependent switching. [22] investigated stochastic control of mean-field type for continuous-time systems with state-and disturbance-dependent noise. For some results of robust control on this kind of systems, we refer to the monograph [23]. However, most of the results were only concerned with either transient performance or steady state performance.
Motivated by aforementioned discussions, the unified stability and stabilization problems for asymptotic and transient behavior of stochastic systems are considered in this paper. By using stochastic analysis technology, the stability criteria and some stabilizing conditions are obtained. The contributions of this paper lie in the following three aspects: (1) A concept of mean square (γ, α)-stability is first introduced to put asymptotic and transient behavior of Itô stochastic systems into a unified framework, and two stability criteria are obtained. (2) By utilizing an auxiliary definition of mean square (γ, T)-stability, the relations among (γ, α)-stability, (γ, T)-stability and FTS are established. (3) On the basis of stability criteria obtained, two kinds of unified stabilization controllers (state and output feedback controllers) are designed to make closed-loop systems have the desired transient and steady performance, and a numerical algorithm is given to obtain the relation between γmin and α.
This paper is organized as follows: Section 2 gives a mean square (γ, α)-stability and its relation to mean square (γ, T)-stability and FTS. Section 3 provides several stability conditions for mean square (γ, α)-stability and mean square (γ, T)-stability and analyzes the relations among these stability conditions. Section 4 is to design state and output feedback mean square (γ, α)-stabilization controllers. A numerical algorithm is given in Section 5. Section 6 employs an example to illustrate the results. Section 7 gives the conclusion.
Notations: XT stands for transpose of a matrix X. The X > 0 means that X is positive-definite. In × n stands for n × n identity matrix. λmax (X)(λmin (X)) represents the maximum(minimum) eigenvalue of a matrix X. denotes the operator of the mathematical expectation. tr(X) is the trace of a matrix X. The notation denotes Euclidian 2-norm of a vector x and ‖x‖Q denotes . The asterisk “*” represents the symmetry term in a matrix and diag{⋅⋅⋅} represents a block-diagonal matrix. The shorthand “wrt” is an abbreviation of “with respect to”. k(X) = λmax (X)/λmin (X) stands for condition number of a X > 0.
Section snippets
Preliminaries and definitions
Consider the following linear Itô stochastic system where is the system state, A, are constant matrices and x0 is the initial state. Without loss of generality, we assume w(t) to be one-dimensional standard Wiener process defined on the probability space (Ω, P) with = σ{w(s): 0 ≤ s ≤ t}.
Next, a definition of the unified description for asymptotic behavior and transient behavior of system (1) is given.
Definition 1 Given the constants γ > 1 and α
Mean square-(γ, α)-stability analysis
This section first gives a sufficient condition for mean square (γ, α)-stability and a sufficient condition for mean square (γ, T)-stability. In regards to stability conditions, the relations among mean square (γ, α)-stability, mean square (γ, T)-stability and FTSS are shown.
Theorem 2 Given γ > 1 and α > 0, system (1) is mean square (γ, α)-stable if there exists a matrix Q > 0 such that the following inequalities hold:
Proof Let Q > 0 such that (6) and (7) are satisfied. Defining
Mean square (γ, α)-stabilization
Consider the following linear stochastic controlled system: where and are the control input, measurement output, respectively. B, and are constant matrices.
Based on Definition 1, mean square-(γ, α)-stabilization can be given as follows.
Definition 4 Given the constants γ > 1 and α > 0, system (25) is called mean square (γ, α)-stabilizable if there exists a feedback controller u*(t) such that the closed-loop
Numerical algorithm
In this section, an algorithm is given to achieve the relation between the minimum value of γ and α. The following analysis and algorithm for Theorem 5 are first given. Similar algorithm can be applied to Theorem 6.
Analysis: By analyzing (27) and (28), we find that if (27) and (28) have feasible solutions wrt X, Y, σ1 and σ2 for given γ and α, then γmin is existent and αmin = 0. So, we first fix α1 = 0 and search for γmin from 1, and then we fix α2 and search for γmin from 1 again, and so
Simulation results
This section provides an illustrative example to demonstrate the effectiveness of the proposed results. Consider system (25) with parameters as follows: Let the scalars γ = 5 and α = 0.2, x(0) = . Using Euler–Maruyama method to simulate the standard Brownian motion, one can obtain the curves of ‖x(t)‖2 (30 curves) and of system (1) in Fig. 1. From Fig. 1, it can be seen that the system (1) is mean square
Conclusion
In this study, a unified framework for asymptotic and transient behavior of stochastic systems has been established. Two criteria to test mean square (γ, α)-stability of a system have been obtained. Furthermore, state feedback and output feedback controllers have been designed for the mean square (γ, α)-stabilization of this class of systems, respectively. In addition, the relations among mean square (γ, α)-stability, mean square (γ, T)-stability and finite-time stability have been established.
Acknowledgments
This work was supported in part by the Natural Science Foundation of CQ under Grant CSTC 2014J-CYJA40004, and the Natural Science Foundation of China under Grant 11471061. Also, the work of Z. Yan was supported by the National Natural Science Foundation of China under Grant no. 61403221 and Project funded by China Postdoctoral Science Foundation (2017M610425).
References (25)
- et al.
Finite-time stabilization of linear time-varying continuous systems
IEEE Trans. Autom. Control
(2009) - et al.
Finite-time control of linear stochastic systems
Int. J. Innov. Comput. Inf. Control
(2008) - et al.
Finite-time tracking control for Markov jump repeated scalar nonlinear systems with partly usable model information
Information Sciences
(2016) - et al.
Finite-time asynchronous filtering for discrete-time Markov jump systems over a lossy network
Int. J. Robust Nonlinear Control
(2016) Stochastic Differential Equations: An Introduction with Applications
(2010)- et al.
Statistical physics of human cooperation
Phys. Rep.
(2017) - et al.
On stabilizability and exact observability of stochastic systems with their applications
Automatica
(2004) - et al.
Stability of stochastic nonlinear systems with state-dependent switching
IEEE Trans. Autom. Control
(2013) - et al.
Mathematical methods in robust control of linear stochastic systems
(2006) - et al.
Linear matrix inequalities in systems and control theory
(1994)
Power System Control and Stability
Short time stability in linear time-varying systems
Proceedings of the IRE International Convention Record
Cited by (15)
Integral sliding mode control for uncertain impulsive stochastic systems
2024, European Journal of ControlObserver-based finite-time asynchronous sliding mode control for Markov jump systems with time-varying delay
2022, Journal of the Franklin InstituteH<inf>∞</inf>control for Poisson-driven stochastic systems
2021, Applied Mathematics and ComputationStabilization for hybrid stochastic systems by aperiodically intermittent control
2021, Nonlinear Analysis: Hybrid SystemsCitation Excerpt :On the one hand, one of the important issues in the study of stabilization of stochastic differential equations with Markovian switching is the design of appropriate controller. In recent years, the stabilization of hybrid stochastic systems has attracted great attention [15–22]. Various controllers have been developed to make an unstable system to be stable one such as feedback controller, pinning controller, impulsive controller, fuzzy controller, intermittent controller and so on.
Admissibility analysis of stochastic singular systems with Poisson switching
2020, Applied Mathematics and ComputationCitation Excerpt :In [20], the problem of reliable dissipative control for singular networked cascade control systems with cyber-attacks was investigated. Recent years have witnessed the rapid development of the analysis of stochastic systems due to their extensive applications in various practical systems [21–23] and important progresses have been achieved [24,25]. Motivated by this, deterministic singular systems have also been generalized to stochastic cases and interesting results are reported for stochastic singular systems [26,27].
Sliding mode control for Markovian jump repeated scalar nonlinear systems with packet dropouts: The uncertain occurrence probabilities case
2019, Applied Mathematics and ComputationCitation Excerpt :Markovian jump systems (MJSs) can be seen as a special class of stochastic dynamical systems with the transitions between different systems governed by a Markov process [17–21]. Such MJSs have received increasing investigation attention in recent years due to the convenience characterization of the practical systems with structures changed randomly [22,23]. Accordingly, a large number of interesting results for MJSs have been reported to meet various requirements, such as robust stabilization and stability analysis [24,25], filter and controller design [14,26,27].