Adaptive Tracking Control for Stochastic Nonlinear Systems with Full-State Constraints and Unknown Covariance Noise

https://doi.org/10.1016/j.amc.2020.125397Get rights and content

Highlights

  • An enhanced adaptive control method is extended to more general constrained stochastic nonlinear system with unknown covariance noise and parameters.

  • A distinctive feature is that the adaptive technique is introduced to handle the unknown covariance noise and parameters simultaneously.

  • An adaptive state-feedback controller is constructed by using backstepping, barrier Lyapunov function (BLF) and adaptive technique.

  • The property of uniformly ultimately bound is obtained.

Abstract

This paper is devoted to the adaptive state-feedback tracking control for stochastic nonlinear systems disturbed by unknown covariance noise under the condition of full-state constraints and parametric uncertainties. Different from the related literatures, nonlinear functions in the diffusion terms are allowed to be unknown in this paper. The parametric uncertainties and unknown covariance noise are compensated with the aid of adaptive control design. By combining the backstepping technique with barrier Lyapunov function (BLF) in a unified framework, the full-state constraints can be dealt. Then, an adaptive state-feedback controller is constructed, which guarantees all the signals in the closed-loop system are uniformly ultimately bounded, the system states remain in the defined compact sets and the output tracks the reference signal well. Finally, stochastic noise is introduced to establish a stochastic simple pendulum system to show the effectiveness of the proposed controller.

Introduction

As is well-known that nonlinear system control plays a crucial role in practical engineering field such as the single-link robot, the tunnel diode circuit, the chemical reactor systems, and the complex networked systems [1], [2], [3], etc.. Based on nonlinear stability theory [4], Lyapunov function control method [5], the backstepping technique [6], and other control methods, numerous control results have been emerged for nonlinear systems in [7], [8], [9], [10], [11], [12], [13], [14] and the references therein. On the other hand, stochastic disturbances and observation noise including the parameter perturbations, system’s control input variations and stochastic errors, widely occur in physical systems and complex networked systems [15], [16], [17], [18], [19], [20], [21]. This makes practical nonlinear systems sometimes can be modeled by stochastic nonlinear differential equations and called as stochastic nonlinear systems. By generalizing the backstepping technique to stochastic nonlinear systems [22] and using the basic stochastic stability theory [23], [24], the controller design for different classes of stochastic nonlinear systems are fully considered; e.g., large-scale stochastic nonlinear systems [25], stochastic nonlinear systems with high-order [26], [27], stochastic nonlinear systems with time delays [28], [29], [30], stochastic switched nonlinear systems [31], [32] and inverse dynamic stochastic nonlinear systems [33], etc.

Note that constraints with the form physical stoppage, performance and safety specifications frequently appear in practical systems, which sometimes deteriorates system performance and stability. It is proven that the barrier Lyapunov function (BLF) proposed by [34] is an effective tool since it can yield a value that approaches infinity while its arguments approach some limits. Since then, BLF together with backstepping technique, neural networks or fuzzy approximation methods has witnessed remarkable control results for deterministic nonlinear systems; see, e.g., [35], [36], [37], [38] and the related references. In contrast to nonlinear systems in deterministic forms, the full-state constrained stochastic nonlinear systems are more difficult to be controlled and the design procedure is more complicated. Until now, for stochastic nonlinear systems with full-state constraints, the related design results are very few and remain to be further addressed. This enlightens us to further take stochastic nonlinear system into accountdxi=ηixi+1dt+θiTfi(x¯i)dt+gi(x¯i)Σdω,i=1,2,,n1,dxn=ηnudt+θnTfn(x¯n)dt+gn(x¯n)Σdω,y=x1,where xiR, yR and uR are the system measurable states, output and control input, respectively, which are constrained within compact sets as |xi| < kbi with kbi being known constants; x¯i=(x1,x2,,xi)T; constant vectors θ1Rm,θ2Rm,,θnRm are unknown and called as the parametric uncertainties; fi(x¯i)Rm denotes the known smooth functions with fi(0)=0; the unknown smooth function gi(x¯i)R1×r satisfies gi(0)=0; nonnegative definite matrices Σ:R+Rr×r is an unknown bounded function and the infinitesimal covariance ΣΣT is a function of Σdω; ω which is defined on a complete probability space, is an r-dimensional standard Wiener process with incremental covariance ΣΣTdt, this means E(dωdωT)=ΣΣTdt.

When ΣΣT=I and |xi| < kbi, some related results have been obtained for system (1). With the help of BLF, backstepping, fuzzy and neural network approximation methods, [39], [40], [41] investigated the feedback control problems for system (1) in various structures. But [39], [40], [41] did not take the parametric uncertainties into account. Note that parametric uncertainties exhibit more complex dynamics commonly exist, such as the machines with friction and the biochemical processes. It is significant to consider the parametric uncertainties when stabilizing the stochastic nonlinear systems. In [42], by constructing BLF in symmetric and asymmetric forms and using the adaptive design, a control scheme was presented for stochastic nonlinear systems with unknown parameters. For stochastic nonlinear systems with parametric uncertainties, based on the tan-type BLF and finite-time backstepping design, an adaptive finite-time controller was constructed in [43].

When ΣΣT ≠ I, the stabilization problem for system (1) has been studied in [44], [45], [46] by using stochastic controller design methods and theory without considering the full-state constraints |xi| < kbi. To the authors’ knowledge, the control for full-state constrained stochastic nonlinear systems with ΣΣT ≠ I is still unsolved, not mention to considering the parametric uncertainties.

In this paper, we focus on the mentioned control problem that is unsolved. The main work and distinctive features are listed as follows.

  • This note makes the first try to address the adaptive control problem of system (1) in more general form containing full-state constraints, uncertain parameters and unknown covariance noise.

  • By using the BLF and adaptive control technique to overcome the negative effect brought by state constraints, parametric uncertainties and unknown covariance noise, an adaptive state-feedback controller is designed. It can be shown that the closed-loop system is uniformly ultimately bounded (UUB), the states are not violated and the system output can be used to track a reference signal.

  • Compared with the related references [42]-[43], the distinctive features of this paper are three folds: the considered system is more general with the covariance noise being considered; the exactly known knowledge of gi is removed; a practical stochastic simple pendulum system instead of a completely numerical example is skillfully established and used to illustrate the effectiveness of the proposed schemes.

The rest of this paper is organized as follows. Section 2 gives some mathematical preliminaries. In Section 3, an adaptive controller is given. Section 4 shows the stability analysis of the controller. A simulation example is provided in Section 5. This is followed by Section 6 to conclude this paper.

Notations: Throughout the whole paper, R+ denotes all the non-negative real numbers set; Rn stands for n-dimensional Euclidean space; Ci represents the family of all the functions with continuous ith partial derivations; XT denotes the transpose of a given vector or matrix X and ‖ · ‖ denotes its Euclidean norm with Tr{X} being its trace when X is square; | · | is the absolute value of a number or a function; a function γ:R+R+ is called as a class K function if it is continuous, strictly increasing and γ(0)=0; K is the set of all functions which are of class K and γ(x) → ∞ if x → ∞.

Section snippets

Mathematical Preliminaries

For the normal stochastic nonlinear systemdx=f(x)dt+g(x)dω,x(0)Rn,where the state vector xRn; ω is defined as those in (1); f(x): RnRn and g: RnRn×r satisfying f(0)=0 and g(0)=0 are locally Lipschitz.

Definition 1 [24]. For any C2 function V(x)Rn along system (2), denote the differential operator of V as LV, then, LV along system (2) isLV(x)=V(x)xf(x)+12Tr{gT(x)2V(x)x2g(x)},where the term 12Tr{gT2V(x)x2g} is called the Herssian term.

Lemma 1

[25]. Consider system (2), and assume there exist a

Design of Adaptive State-Feedback Controller

This section gives the controller design procedure by the backstepping technique, which is based on the state coordinate transformation with the formz1=x1yr,zi=xiαi,i=2,3,,n,where yr is the reference signal satisfying Assumption 1 and α2, α3, ⋅⋅⋅, αn are virtual control laws need to be designed. In addition, one further defines the adaptive parameter Θ with the following form to handle the unknown covariance noiseΘmax{ΣΣT2,ΣΣT43}.

Then, using (1)-(4) and the Itô’s differentiation

Stability Analysis

The theorem contains the main results is concluded as follows.

Theorem 1

Considering system (1) under Assumptions 1-2, if the virtual control laws are designed as (12), (20), (30), the adaptive laws are given by (13), (21), (31) and (34)-(35), the actual controller is constructed as (33) and the initial conditions are chosen in zi(0)Ωz0:={Xn(0)R2n+2:|zi(0)|<kci}, where Xn(0)=(z¯n(0),θ^¯s,n(0),Θ^(0),yr(0))T with z¯n=(z1,z2,,zn)T and θ^¯s,n=(θ^s,1,θ^s,2,,θ^s,n)T, then, it holds (i). all the signals in

A Simulation Example

Take the simple pendulum system shown in [6] into consideration. The equation of motion in the tangential direction is given bymlθ¨=mgsin(θ)klθ˙,where l denotes the length of the rod; m denotes the mass of the bob; θ is the angle subtended by the rod and the vertical axis through the pivot point; k is a coefficient of friction. Applying a torque T and viewed it as a control input. Then, define x1=θ, x2=θ˙ and u=T, one hasx˙1=x2,x˙2=1ml2uglsin(x1)kmx2.

Following [46], the Gaussian white noise

Conclusions

This paper has studied the state-feedback control for full-state constrained stochastic nonlinear systems in more general forms. The BLF is used to handle the full-state constraints and the adaptive control technique is applied to deal with the covariance noise and parametric parameters. Through the backstepping procedure, an adaptive controller is designed and guarantees the closed-loop system to be UUB, the tracking error is within a compact set of the origin and the full-states are not

Conflict of Interest

We declare that none of the authors have a conflict of interests.

Acknowledges

This work was supported in part by the Project funded by China Postdoctoral Science Foundation under grant 2019M661692; in part by NSFC of China under Grant 61673215, Grant 61473151, Grant 61573172, Grant 61977037 and Grant 61703059.

References (46)

  • Y.J. Liu et al.

    Barrier lyapunov functions-based adaptive control for a class of nonlinear pure-feedback systems with full state constraints

    Automatica

    (2016)
  • W. Xiao et al.

    Adaptive fuzzy control for pure-feedback systems with full state constraints and unknown nonlinear dead zone

    Applied. Math. Comput.

    (2019)
  • M. Xia et al.

    Adaptive neural dynamic surface control for full state constrained stochastic nonlinear systems with unmodeled dynamics

    J. Frankl. Inst.

    (2019)
  • Y.J. Liu et al.

    Adaptive control-based barrier lyapunov functions for a class of stochastic nonlinear systems with full state constraints

    Automatica

    (2018)
  • J. Zhang et al.

    Finite-time tracking control for stochastic nonlinear systems with full state constraints

    Appl. Math. Comput.

    (2018)
  • H. Min et al.

    Adaptive finite-time control for stochastic nonlinear systems subject to unknown covariance noise

    J. Frankl. Inst.

    (2018)
  • H. Li et al.

    Event-triggered communication and data rate constraint for distributed optimization of multi-agent systems

    IEEE Trans. Syst. Man Cybern. Syst.

    (2017)
  • H. Li et al.

    Distributed consensus optimization in multi-agent networks with time-varying directed topologies and quantized communication

    IEEE Trans. Cybern.

    (2017)
  • J. Zhou et al.

    Weighted h consensus design for stochastic multi-agent systems subject to external disturbances and ADT switching topologies

    Nonlin. Dyn.

    (2019)
  • H.K. Khalil

    Nonlinear Systems, Third Edition

    (2002)
  • E.D. Sontag

    Input to state stability: Basic concepts and results

  • M. Krstić et al.

    Nonlinear and Adaptive Control Design

    (1995)
  • C.J. Qian et al.

    Practical output tracking of nonlinear systems with uncontrollable unstable linearization

    IEEE Trans. Autom. Control

    (2002)
  • Cited by (19)

    • Prescribed tracking error fixed-time control of stochastic nonlinear systems

      2022, Chaos, Solitons and Fractals
      Citation Excerpt :

      As the existence of stochastic disturbance may lead to system instability, tons of attention has been paid to the adaptive control of stochastic system, and some effective schemes have been acquired in [1–9].

    • Decentralized stabilization of large-scale stochastic nonlinear systems with time-varying powers

      2022, Applied Mathematics and Computation
      Citation Excerpt :

      For the systems with time-varying powers, many important issues are still open and worth, e.g., how to achieve decentralized stabilization when Assumption 3 is removed, how to generalize the results to more general system [36–42].

    View all citing articles on Scopus
    View full text