Elsevier

ISA Transactions

Volume 125, June 2022, Pages 110-118
ISA Transactions

Research article
Fuzzy adaptive finite-time output feedback control of stochastic nonlinear systems

https://doi.org/10.1016/j.isatra.2021.06.029Get rights and content

Highlights

  • The dynamic surface method solves the problem of “computational explosion”.

  • A fuzzy state observer is constructed to estimate the unmeasured state.

  • A new output feedback controller is designed.

Abstract

An adaptive finite-time approach to the feedback control of stochastic nonlinear systems is presented. The fuzzy logic system (FLS) and a state observer are used to estimate the uncertain function and unmeasured state of the controlled system, respectively. A dynamic surface control (DSC) scheme is employed to deal with the “computational explosion” problem, which is inherent in traditional backstepping methods since the repetitive calculation of the derivatives of virtual control signals is avoided. A new output feedback controller is developed to guarantee that all the signals of the controlled system are bounded within a finite time range and the tracking deviation can converge to an arbitrarily small residual set within finite time. Simulations confirm the analytical and theoretical results of the presented algorithm.

Introduction

In many practical systems, there will be random disturbances, which may be the source of system instability. How to design and analyze the controller in a stochastic system has become particularly important and significant progress has been made in this field [1], [2], [3], [4], [5], [6]. In [7], the problem of robust sliding mode control was studied for stochastic systems with randomly occurring nonlinearities, time delays and uncertain discrete-time. The moment stability and sample-path stability were analyzed for switched stochastic systems in [8]. A unified optimal and exponentially stable filter was developed in [9] for stochastic systems with linear discrete-time. In [10], the result of the stability of the linear stochastic system with state and control related noise in a finite time was reported, considering that the filter simultaneously estimated the state and the unknown in the sense of unbiased minimum variance input without any assumptions of directly feeding the matrix.

The backstepping control algorithm is a popular design tool for system stability analysis. The application of backstepping technology has solved the stability problems of various types of nonlinear systems with random disturbance and there have been many remarkable results [11], [12], [13], [14], [15]. The authors have achieved asymptotic tracking and interference attenuation for stochastic nonlinear strict-feedback systems described in the form of strict parameter feedback and affected by other external interference inputs in [16] and this was the first time that the backstepping scheme was applied to stochastic systems. A nonlinear pure feedback system with random disturbance and unknown dead zone was considered, and its tracking control problem was solved in [17] and the proposed method ensured that the system achieves semi-globally, uniformly bounded in probability. The adaptive approach to the tracking control problems the output feedback canonical systems corrupted by Wiener noise with unknown covariance were addressed [18]. In the traditional backstepping control method, repeated derivation of the virtual function may cause “computational explosion”, which is also its inherent problem. In order to solve the problem of “computational explosion”, the algorithm of dynamic surface control (DSC) was considered firstly. The purpose of this algorithm is to estimate the derivative of the intermediate variable function by using the output of the filter. Subsequently, there have been many promising results [19], [20], [21], [22]. In [23], an adaptive DSC method was reported based on neural networks for nonaffine nonlinear systems with unknown time delay and input delay nonlinearity. The trajectory tracking problem of marine vehicles with unknown time-varying disturbance, actuator rate limit and input saturation was studied and an n-DOF (degree-of-freedom) nonlinear control law was developed by using disturbance observer and nonlinear DSC technology in [24]. In [25], the problem of an adaptive neural tracking control was addressed for uncertain nonlinear systems with time-delay, random disturbance, input and output constraints. Because it does not need to calculate a large number of derivatives of intermediate variable signals, the proposed DSC control method overcomes the inherent computational complexity of classic backstepping control methods.

Compared with asymptotic stability, it provides fast response speed, high tracking accuracy and strong anti-jamming ability. The finite-time convergence problem has become a meaningful and interesting topic and many excellent results have been achieved [26], [27], [28], [29], [30], [31], [32]. The result of the optimal finite-time stabilization was considered for nonlinear systems in [33]. In [34] the tracking problem was solved for the distributed Euler–Lagrange dynamic network and a finite-time algorithm was developed. For non-strictly feedback nonlinear switching systems with incalculable states, arbitrary switching, and unknown functions composed of the entire state, the problem of finite-time switching control was studied in [35]. For the non-linear quantization system in [36], whose state could not be measured, the finite-time quantitative feedback control was studied for the first time, which solved the problem of finite-time tracking. The application of the finite-time method improves the tracking performance of the system. In the process of system stability analysis, the handling of unknown items has always attracted attention. For unknown functions, two intelligent methods, fuzzy logic system (FLS) [19], [37] and neural networks (NNs) [38], [39], [40], [41], were widely used. For the processing of unknown states, the construction of the state observer was particularly important [42], [43]. However, to our knowledge, the stability analysis of stochastic nonlinear systems with unmeasured states has not yielded results in a finite-time frame. This is the motivation for this work.

In this paper, we study the finite-time control problem for a class of stochastic nonlinear systems with unknown states. The fuzzy logic system is adopted to estimate the unknown function and the designed state observer is used to estimate the unmeasured state. The application of DSC in backstepping control alleviates the calculation problem of virtual control function. The finite-time method improves both the tracking speed and convergence performance. The novelties of this study are listed as follows:

(1) A DSC method is introduced to solve “computational explosion” problem caused by the repeated derivation of intermediate variable function. The DSC method is applied in each step of backstepping scheme and the output of the first-order filter is used to approximate the derivative of the virtual control function. Compared with traditional schemes, the uniqueness of this scheme is based on the finite-time frame in that it uses DSC technology in the backstepping control step.

(2) Nonlinear systems with random disturbance and unknown state are considered within a finite-time frame, which is different from [44]. Constructing a fuzzy state observer can estimate the unmeasured state. A new output feedback controller is developed to guarantee that all the signals of the closed-loop system are bounded within a finite time and the error of tracking is able to converge to arbitrarily small residual sets within a finite time.

Section snippets

Preliminaries and system description

The form of a class of stochastic nonlinear systems is given as follows: dxi=(fi(Xi)+xi+1)dt+gi(Xi)dω(1in1)dxn=(fn(X)+u)dt+gn(X)dωy=x1where the state vector is X=[x1,x2,,xn]T, u represents the control input and y is the output. fi(Xi) and gi(Xi),i=1,2,,n denote continuous, smooth and uncertain functions, where Xi=[x1,x2,,xi]. ω is an r-dimensional standard Brownian motion. ω is defined on a complete probability space with the incremental covariance E{dωdωT}=ξ(t)ξ(t)Tdt, ξ(t) is an

Fuzzy state observer design

Define Wi=θi, i=1,2,,n, θi is an unknown parameter vector, so from Lemma 5, we can know fˆi(Xˆi|θˆi)=θˆiTSi(Xˆi),fˆi(Xˆi|θi)=θiTSi(Xˆi), where θˆi is the estimation of θi, the estimated error is θ̃i=θiθˆi, we have θi=argminθˆiΩi[supXˆiUi|fˆi(Xˆi|θˆi)fi(Xˆi)|]where Ωi is θˆi compact set and Ui is Xˆi compact set. The fuzzy minimum estimation error is defined as follows δi=fi(Xˆi)fˆi(Xˆi|θi)where |δi|<ɛi with ɛi is an unknown positive constant. The entire system (1) can be expressed as: dX=(

Main results

In this section, an algorithm based on DSC and backstepping technique is presented. We first give the coordinate transformation as follows z1=x1xd,zi=xˆiβi,i=2,,nwhere xd is the reference output signal and βi represents the output signal of the filter.

βi is the output of the first-order filter when αi1 is the input, αi1 is an virtual control function and τi is a time constant. βi can be introduced as τiβ̇i+βi=αi1,βi(0)=αi1(0),i=2,,n.with the filter errors χi=βiαi1.

Step 1: According to

Simulation examples

In this section, we present two simulation examples to illustrate the effectiveness of our proposed control scheme.

(i) Numerical Example: a strict-feedback stochastic nonlinear system is given: dx1=(f1(X1)+x2)dt+g1T(X1)dωdx2=(u+f2(X2))dt+g2T(X2)dωy=x1where x1 and x2 are the states, the output is y and u is control input. f1=0.1×x1sin(x1),f2=x1x22,g1=0.01×x13 and g2=0.01×x1x22.

The form of the state observer is given as follows xˆ̇1=xˆ2+θˆ1S1T(Xˆ1)+k1(yyˆ)xˆ̇2=u+θˆ2S2T(Xˆ2)+k2(yyˆ).It is

Conclusion

An adaptive finite-time algorithm is presented for nonlinear systems with random disturbance and unknown state. The FLS scheme is adopted to estimate unknown functions while the unmeasured state is estimated by using a designed state observer. By applying the dynamic surface control method, the significant computation load due to the multiple derivations of the intermediate variable function is solved. The development of a finite-time output feedback controller ensures that all the signals of

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work was supported in part by the Fundamental Research Funds for the Central Universities 2021YJS140, and in part by the National Science Foundation of China under projects 61761166011, 61773072 and 61773051, and in part by the Education Department of Liaoning Province under the general project research under Grant No. LJ2020001.

References (52)

  • LiJ. et al.

    Adaptive nn output-feedback stabilization for a class of stochastic nonlinear strict-feedback systems

    ISA Trans

    (2009)
  • ZhuQ. et al.

    Adaptive neural network output feedback control for stochastic nonlinear systems with full state constraints

    ISA Trans

    (2020)
  • HuangX. et al.

    Global finite-time stabilization of a class of uncertain nonlinear systems

    Automatica

    (2005)
  • ZhangH. et al.

    Command filter based adaptive fuzzy finite-time control for a class of uncertain nonlinear systems with hysteresis

    IEEE Trans Fuzzy Syst

    (2020)
  • WangH. et al.

    Neural-based adaptive output-feedback control for a class of nonstrict-feedback stochastic nonlinear systems

    IEEE Trans Cybern

    (2015)
  • ZhaoxuY. et al.

    Razumikhin-Nussbaum-lemma-based adaptive neural control for uncertain stochastic pure-feedback nonlinear systems with time-varying delays

    Internat J Robust Nonlinear Control

    (2012)
  • WangH. et al.

    Adaptive neural tracking control for a class of nonstrict-feedback stochastic nonlinear systems with unknown backlash-like hysteresis

    IEEE Trans Neural Netw Learn Syst

    (2014)
  • HuJ.

    A delay fractioning approach to robust sliding mode control for discrete-time stochastic systems with randomly occurring non-linearities

    IMA J Math Control Inform

    (2011)
  • WeiF. et al.

    Stability analysis of switched stochastic systems

    Automatica

    (2011)
  • YanZ. et al.

    Finite-time stability and stabilization of linear it stochastic systems with state and control-dependent noise

    Asian J Control

    (2013)
  • JingZ. et al.

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

    Appl Math Comput

    (2018)
  • WangH. et al.

    Direct adaptive neural control for strict-feedback stochastic nonlinear systems

    Nonlinear Dynam

    (2012)
  • LiuS.J. et al.

    Output-feedback control of a class of stochastic nonlinear systems with linearly bounded unmeasurable states

    Internat J Robust Nonlinear Control

    (2010)
  • PanZ. et al.

    Adaptive controller design for tracking and disturbance attenuation in parametric strict-feedback nonlinear systems

    IEEE Trans Automat Control

    (1998)
  • WangH. et al.

    Direct adaptive neural tracking control for a class of stochastic pure-feedback nonlinear systems with unknown dead-zone

    Internat J Adapt Control Signal Process

    (2013)
  • JiH. et al.

    Adaptive output-feedback tracking of stochastic nonlinear systems

    IEEE Trans Automat Control

    (2006)
  • Cited by (20)

    • Observer and fault reconstruction-based control design for interval type-2 fuzzy systems

      2023, Communications in Nonlinear Science and Numerical Simulation
    • Command filter-based adaptive fuzzy decentralized control for stochastic nonlinear large-scale systems with saturation constraints and unknown disturbance

      2023, ISA Transactions
      Citation Excerpt :

      This constraint restraints the applied range of backstepping technique. As a replace method, many intelligent control technologies have been achieved, such as neural networks (NNs) and FLSs [28–34]. Because of their strong approximation ability to continuous nonlinear functions [35,36], the difficulty of nonlinear systems with unstructured uncertainties can be resolved.

    View all citing articles on Scopus
    View full text