Sage Journals: Discover world-class research

Abstract

In this article, an adaptive neural tracking controller is designed for near-space vehicles with stochastic disturbances and unknown parametric uncertainties. Based on the great nonlinear function approximation capability of neural networks, the unknown system uncertainties are tackled using the radial basis function neural networks. Furthermore, on the basis of stochastic Lyapunov stability theory, an adaptive tracking control scheme is developed for near-space vehicle which can guarantee the closed-loop system stability. Under the developed adaptive neural control scheme, all closed-loop system signals are bounded in the sense of probability, and the tracking error converges to a small neighborhood of the origin. Finally, simulation results are provided to illustrate the proposed adaptive neural control scheme that can guarantee the satisfactory tracking performance for the attitude motion of the near-space vehicle with stochastic disturbances.

Keywords

Near-space vehicles stochastic disturbances neural network adaptive control backstepping

Introduction

Near-space vehicles (NSVs) are novel aerospace vehicles that have attracted more and more attention. With the advantages of traditional aircrafts and spacecrafts, such as high mobility and wide flight envelope,^1

–4 it has raised concerns on military and civilian application. As a kind of flight control systems, NSV has the complicated characteristics, such as strong nonlinearity, strong coupling, system uncertainties, special working environment, and multiple mission modes. The robust flight control of NSV becomes a challenging and valuable work, and some control approaches have been proposed in the literature.^{5

–15} Meanwhile, the control design as one of the most important research topics in NSV, the tracking control problem has been extensively studied. The novel global neural dynamic surface tracking control scheme is presented for hypersonic flight vehicle in the study by Xu et al.¹⁶ The guaranteed transient performance-based attitude control scheme was proposed for NSV with input saturation by Chen et al.¹⁷ In the study by Chen et al.,¹⁸ a robust sliding mode control strategy was presented for attitude dynamics of NSV. In the study by Cheng et al.,¹⁹ an online support vector regression compensated nonlinear-generalized predictive control (NGPC) system was presented for hypersonic vehicles.

As is well known, the NSV is a multiple-input multiple-output (MIMO) nonlinear system subject to external deterministic disturbances which will further increase the design difficulty of the reliable and accurate flight control scheme. Over the past few decades, the disturbance observer–based control schemes have been well studied in the literature.^{20

–30} To efficiently handle the unknown disturbance based on the dynamic information of disturbance, the adaptive dynamic surface control scheme based on disturbance observer was proposed for the NSV with input saturation by Chen and Yu.³¹ In the study by Chen et al.,³² on the basis of terminal sliding mode technique and disturbance observer method, an anti-disturbance control scheme was developed for the hypersonic flight vehicles with input saturation. In the study by Cheng et al.,³³ an NGPC strategy based on sliding mode disturbance observer (SMDO) was designed for the hypersonic NSV attitude control system. However, stochastic disturbances always exist in practical systems which need to be further considered. Especially, the flight attitude control of NSV will become more complicated if above mentioned aspects are simultaneously considered.

Recently, some significant results on backstepping-based control scheme^{34

–41} for deterministic nonlinear systems have been generalized to stochastic nonlinear systems. For instance, by introducing a quartic Lyapunov function, a backstepping design scheme was proposed for a class of strict-feedback stochastic nonlinear systems by Deng et al.⁴² Based on the great nonlinear function approximation capability of neural networks (NNs) or fuzzy logic systems, by combining adaptive fuzzy control or adaptive neural control with backstepping technique, many adaptive control schemes can be used to efficiently handle the uncertainties of stochastic nonlinear systems. Furthermore, some meaningful results were presented for stochastic nonlinear systems on the basis of the quartic Lyapunov function in the literature.^{43

–55} It is well known that mechanical systems are often subjected to stochastic disturbances, which significantly affect the performance. By reasonably introducing random noise, a method to construct stochastic Lagrangian control systems was given and an adaptive tracking controller was designed by Cui et al.⁵⁶ Using geometric stochastic feedback control, a robust asymptotic stabilization of rigid body attitude dynamics with stochastic input torque was studied by Samiei et al.⁵⁷ In earlier study,⁵⁸ a new robust stochastic control methodology was developed for unmanned aerial vehicles. In a realistic environment, according to the characteristics of the NSV, stochastic components and random noises are included in the models which will lead to the considerable difficulty in the design stage of the flight control system for the NSV with stochastic disturbances.

Motivated by above analysis, in this article, using backstepping technology, an adaptive neural tracking control scheme will be developed for the NSV with stochastic disturbances, unknown parametric uncertainties, and external disturbances. This article is organized as follows: The stochastic nonlinear control system of NSV with stochastic disturbance and mathematical preliminaries is given in “Problem statements and preliminaries” section. The design and stability analysis of an adaptive neural tracking controller is proposed in “Adaptive neural tracking control design” section. In “Simulation study” section, simulation results are given to demonstrate the effectiveness of our developed robust attitude control approaches. Conclusions are addressed in the last section.

Throughout this article, ∥ ⋅ ∥ stands for the Euclidean norm (or Frobenius norm for a matrix); λ_min(⋅) and λ_max(⋅) denote the minimum and maximum eigenvalues of a matrix, respectively. Let E[⋅] and P{⋅} represent the expectation operator and the probability of a random variable, respectively. $F$ is the σ-algebra of the subsets of sample space Ξ, and P is the probability measure.

Problem statements and preliminaries

Problem formulation and some assumptions

The uncertain attitude dynamics with external disturbance of the NSV can be described as the following MIMO nonlinear system³:

\begin{matrix} \dot{Ω} = F_{1} (Ω) + Δ F_{1} (Ω) + G_{1} (Ω) ω + d_{1} \\ \dot{ω} = F_{2} (Ω, ω) + Δ F_{2} (Ω, ω) + G_{2} (Ω, ω) M_{c} + d_{2} \end{matrix}

where $Ω = [α, β, μ]^{T} \in ℝ^{3}$ is the attitude angle vector representing attack angle, sideslip angle, and roll angle, respectively. $ω = [p, q, r]^{T} \in ℝ^{3}$ is the attitude angular rate vector representing roll angular rate, pitch angular rate, and yaw angular rate, respectively. $M_{c} = [l_{c}, m_{c}, n_{c}]^{T} \in ℝ^{3}$ is the control moment vector representing rolling moment, pitching moment, and yaw moment, respectively. $F_{1} \in ℝ^{3}$ and $F_{2} \in ℝ^{3}$ are the known state function vectors; $G_{1} \in ℝ^{3 \times 3}$ and $G_{2} \in ℝ^{3 \times 3}$ are the known system control matrices which are invertible. $Δ F_{1}, Δ F_{2} \in ℝ^{3}$ are the unknown modeling errors which are smooth functions, and $d_{1}, d_{2} \in ℝ^{3}$ are the unknown external disturbances.

By considering the influence of the stochastic disturbance torque, as shown in the study by Zhu,⁵⁹ the attitude dynamics in the body frame is obtained as follows

\begin{array}{l} \dot{Ω} = F_{1} (Ω) + Δ F_{1} (Ω) + G_{1} (Ω) ω + d_{1}, \\ G_{2}^{- 1} (Ω, ω) \dot{ω} = G_{2}^{- 1} (Ω, ω) (F_{2} (Ω, ω) + Δ F_{2} (Ω, ω)) + M_{c} + G_{2}^{- 1} (Ω, ω) d_{2} + Λ (Ω, ω) ξ \end{array}

where $Λ (Ω, ω) ξ$ is caused by white noise (i.e. stochastic disturbance signal) ξ, and $Λ (Ω, ω) \in ℝ^{3 \times m}$ is a Borel measurable function matrix which is locally bounded and locally Lipschitz continuous, and $Λ (0, 0) = 0$ .

According to Section 5.5 of the reference,⁶⁰ with $\frac{d W}{d t}$ instead of ξ, we obtain the following stochastic differential equation in the Stratonovich sense as

\begin{array}{l} d Ω = (F_{1} (Ω) + Δ F_{1} (Ω) + G_{1} (Ω) ω + d_{1}) d t \\ d ω = (F_{2} (Ω, ω) + Δ F_{2} (Ω, ω) + G_{2} (Ω, ω) M_{c} + d_{2}) d t + G_{2} (Ω, ω) Λ (Ω, ω) \circ d W \end{array}

where W is a m-dimensional independent Wiener process.

To facilitate the controller design for attitude dynamics with stochastic disturbances, the system (3) needs to be transformed into the equivalent $I t \hat{o}$ type stochastic differential equation as follows

\begin{matrix} {[G_{2}^{- 1} (Ω, ω) d ω]}_{i} = {[G_{2}^{- 1} (Ω, ω) (F_{2} (Ω, ω) + Δ F_{2} (Ω, ω)) + M_{c} + G_{2}^{- 1} (Ω, ω) d_{2}]}_{i} d t \\ + \frac{1}{2} \sum_{k = 1}^{3} \sum_{j = 1}^{m} Λ_{k j} \frac{\partial Λ_{i j}}{\partial ω_{k}} d t + \sum_{j = 1}^{m} Λ_{i j} d W_{j} (t) \end{matrix}

where ${[\cdot]}_{i}, i = 1, 2, 3$ denotes the ith element of a vector, and $\frac{1}{2} \sum_{k = 1}^{3} \sum_{j = 1}^{m} Λ_{k j} \frac{\partial Λ_{i j}}{\partial ω_{k}}$ is the Wong–Zakai correction term.⁵¹ Suppose that the power spectral density (PSD) of white noise ξ equals $\frac{1}{2 π} Π$ , which is equivalent to the fact $d W = Π d B$ , where $Π \in ℝ^{m \times m}$ is an unknown positive matrix and B is a m-dimensional independent standard Wiener process. It is a fact that $E {d W d W^{T}} = Π E {d B d B^{T}} Π^{T} = Π Π^{T} d t$ .⁴² For the ease of control design, equation (4) can be represented using the matrix form of an $I t \hat{o}$ type nonlinear stochastic differential equation as

{\begin{array}{l} d Ω = (F_{1} (Ω) + Δ F_{1} (Ω) + G_{1} (Ω) ω + d_{1}) d t \\ d ω = (F_{2} (Ω, ω) + Δ F_{2} (Ω, ω) + G_{2} (Ω, ω) M_{c} + d_{2} + \bar{G} (Ω, ω)) d t + G_{2} (Ω, ω) Λ (Ω, ω) Π d B (t) \\ y = Ω \end{array}

where ${\bar{g}}_{i} = \frac{1}{2} \sum_{k = 1}^{3} \sum_{j = 1}^{m} Λ_{k j} \frac{\partial Λ_{i j}}{\partial ω_{k}}$ , and $\bar{G} (Ω, ω) = G_{2} (Ω, ω) [{\bar{g}}_{1}, {\bar{g}}_{2}, {\bar{g}}_{3}]^{T} \in ℝ^{3}$ is the Wong–Zakai coefficient vector.

The control object of this article is that the output of system (5) can track the desired bounded signal $Ω_{d} = [α_{d}, β_{d}, μ_{d}]^{T}$ under the proposed adaptive neural controller which can ensure all the closed-loop system signals bounded in the sense of probability, and the tracking errors converge to a small neighborhood of the origin. For simplicity, in the following, the state vectors Ω, ω will be omitted from the corresponding functions. To facilitate the adaptive neural controller design for the NSV described by equation (5), the following assumptions are given as follows.

Assumption 1

For the nonlinear attitude dynamics system (5) of NSV with stochastic disturbances, the gain matrices G₁ and G₂ are invertible.³ In addition, there exist unknown positive constants ${\bar{g}}_{i}$ and $g_{i}^{*}$ such that $λ_{\max} (G_{i}^{T} G_{i}) \leq {\bar{g}}_{i}^{2}$ and $∥ G_{i} ∥ \leq g_{i}^{*}, i = 1, 2$ .

Assumption 2

For all t > 0, there exist the unknown positive constants ϖ_i such that $∥ Ω_{d}^{(i)} ∥ \leq ϖ_{i}, i = 1, 2$ .³

Assumption 3

For the time-varying unknown external disturbances d_i(t), there exist unknown positive constants ${\bar{d}}_{i}$ such that $∥ d_{i} ∥ \leq {\bar{d}}_{i}, i = 1, 2$ .

Assumption 4

There exist unknown positive constants $ρ_{1}, ρ_{2}$ and known smooth nonnegative functions $ψ_{1} (Ω, ω)$ , $ψ_{2} (Ω, ω)$ such that $∥ Λ (Ω, ω) ∥ \leq ρ_{1} ψ_{1} (Ω, ω)$ and $∥ \bar{G} (Ω, ω) ∥ \leq g_{2}^{*} ρ_{2} ψ_{2} (Ω, ω)$ .⁵⁶

Mathematical preliminaries

Consider the following stochastic nonlinear system⁶¹

d x = f (x) d t + g (x) d B, \forall x \in R^{n}

where $x \in R^{n}$ is the state of the system, B is a m-dimensional independent standard Wiener process defined on the complete probability space $(Ξ, F, P)$ , and $f (\cdot) : R^{n} \to R^{n}$ and $g (\cdot) : R^{n} \to R^{n \times m}$ are locally Lipschitz functions in x and satisfy f(0) = 0 and g(0) = 0, respectively.

Definition 1

For any given $V (x) \in C^{2}$ , associated with the stochastic differential equation (6), the differential operator L is defined as follows⁶¹

L V = \frac{\partial V}{\partial x} f (x) + \frac{1}{2} Tr {g^{T} (x) \frac{\partial^{2} V}{\partial x^{2}} g (x)}

where $Tr (\cdot)$ is the trace of a matrix, and $\frac{1}{2} Tr {g^{T} (x) \frac{\partial^{2} V}{\partial x^{2}} g (x)}$ is called $I t \hat{o}$ correction term.

Definition 2

The stochastic process $x (t), t \geq 0$ of stochastic system (6) is said to be bounded in probability, if⁶¹

lim_{c \to \infty} sup_{0 \leq t < \infty} P {| | x (t) | | > c} = 0

Definition 3

The stochastic process⁶¹ x(t) of equation (6) is said to be semi-globally uniformly ultimately bounded (SGUUB) in pth moment, if for a given compact set $Ω^{*} \in R^{n}$ and any initial condition $x_{0} = x (t_{0})$ , there exist a constant ε > 0 and a time constant $T = T (ε, x_{0})$ such that $E [| | x (t {) | |}^{p}] < ε$ for all $t > t_{0} + T .$ Especially, when p = 2, it is usually called SGUUB in mean square.

Lemma 1

Consider the stochastic nonlinear system (6).⁴² If there exists a positive definite, radially unbounded, twice continuously differentiable Lyapunov function $V : R^{n} \to R$ and constants $a_{0} > 0, b_{0} \geq 0$ such that

L V (x) \leq - a_{0} V (x) + b_{0}

then, (i) the system has an unique solution almost surely and (ii) the system is bounded in probability.

Lemma 2 (Young’s inequality)

For $\forall (x^{*}, y^{*}) \in R^{2}$ , the following inequality holds⁴⁵

x^{*} y^{*} \leq \frac{ϑ^{s}}{s} | x^{*} |^{s} + \frac{1}{ι ϑ^{ι}} | y^{*} |^{ι}

where $ϑ > 0, s > 1, ι > 1$ and $(s - 1) (ι - 1) = 1.$

Neural networks

The radial basis function NNs (RBFNNs) are used to approximate an unknown continuous function $f (Z) : R^{r} \to R$ , which can be expressed as¹⁷

\hat{f} (Z) = Φ^{T} S (Z)

where $\hat{f} (Z)$ is the approximation of f(Z), and $Z \in Ω_{Z} \subset R^{r}$ is the input vector with r being the NNs input dimension. $Φ = [w_{1}, w_{2}, \dots, w_{l}]^{T} \in R^{l}$ is the weight vector. l > 1 denotes the NN node number. $S (Z) = [s_{1} (Z), s_{2} (Z), \dots, s_{l} (Z {)]}^{T} \in R^{l}$ is the basis function vector with s_i(Z) being chosen as the commonly used Gaussian function of the form

s_{i} (Z) = \exp [- \frac{{(Z - μ_{i})}^{T} (Z - μ_{i})}{η^{2}}]

where $μ_{i} = [μ_{i 1}, \dots, μ_{i r}]^{T}, i = 1, 2, \dots, l$ is the center of the receptive field and η is the width of the Gaussian function. The RBFNNs (equation (11)) can approximate any continuous function f(Z) defined over a compact set $Ω_{Z} \subset R^{r}$ to arbitrary any accuracy δ. Then, there exists an NN $Φ^{*}^{T} S (Z)$ such that

f (Z) = Φ^{*}^{T} S (Z) + δ, \forall Z \in Ω_{Z} \subset R^{q}

where δ is the approximation error and satisfies |δ| ≤ δ* with δ* > 0. Φ* is the optimal weight vector and it is expressed in the following form

Φ^{*} : = \arg min_{Φ \in R^{l}} {sup_{Z \in Ω_{Z}} | f (Z) - Φ^{T} S (Z) |}

Adaptive neural tracking control design

In this section, the backstepping design is based on the following coordinate transformations

z_{1} = Ω - Ω_{d}, z_{2} = ω - υ_{1}

where z₁ and z₂ are called the error variables, and υ₁ is the virtual control law which will be given later.

To begin with the backstepping design procedure, for i = 1, 2, denote $θ_{i} = ∥ Φ_{i}^{*} ∥$ with $Φ_{i}^{*}$ being the optimal weight vector, and $ρ = \max {| | Π {| |}_{F}^{4} ρ_{1}^{4}, g_{2}^{*}^{\frac{4}{3}} ρ_{2}^{\frac{4}{3}}}$ . The estimations of the unknown parameters θ_i and ρ are ${\hat{θ}}_{i}$ and $\hat{ρ}$ , respectively. And the parameter estimation errors are ${\tilde{θ}}_{i} = θ_{i} - {\hat{θ}}_{i}$ and $\tilde{ρ} = ρ - \hat{ρ}$ , respectively.

Step 1

Noting $z_{1} = Ω - Ω_{d}$ , the error dynamic is

$d z_{1} = (F_{1} + Δ F_{1} + G_{1} ω + d_{1} - {\dot{Ω}}_{d}) d t$
16

As shown in “Neural networks” section, the unknown uncertainty ΔF₁ will be approximated using the RBFNNs. Thus, there exists an NN approximation $Φ_{1}^{}^{T} S_{1} (Ω)$ such that

$Δ F_{1} = Φ_{1}^{ T} S_{1} (Ω) + δ_{1}$
17

where the approximation error vector δ₁ satisfies $∥ δ_{1} ∥ \leq δ_{1}^{}$ with $δ_{1}^{} > 0$ .

Substituting equation (17) into equation (16) yields

$\begin{matrix} d z_{1} = (F_{1} + Φ_{1}^{^{T}} S_{1} + δ_{1} + G_{1} ω + d_{1} - {\dot{Ω}}_{d}) d t \\ = (F_{1} + Φ_{1}^{^{T}} S_{1} + G_{1} ω + D_{1} - {\dot{Ω}}_{d}) d t \end{matrix}$
18

According to assumption 3, the compounded disturbance $D_{1} = δ_{1} + d_{1}$ is bounded, that is, there exists a positive constant $ζ_{1} > 0$ such that $∥ D_{1} ∥ \leq ζ_{1} .$ The estimation of the unknown parameter ζ₁ is ${\hat{ζ}}_{1} .$ And the parameter estimation error is ${\tilde{ζ}}_{1} = ζ_{1} - {\hat{ζ}}_{1}$ .

The virtual control law υ₁ is designed as

$υ_{1} = - G_{1}^{- 1} (K_{1} z_{1} + \frac{3 | | G_{1} {| |}^{\frac{4}{3}}}{4 ς^{\frac{4}{3}}} z_{1} + F_{1} - {\dot{y}}_{d} + {\hat{θ}}_{1} \frac{3 ∥ S_{1} ∥^{\frac{4}{3}}}{4 τ_{1}^{\frac{4}{3}}} z_{1} + {\hat{ζ}}_{1} \frac{3}{4 ς_{1}^{\frac{4}{3}}} z_{1})$
19

where $K_{1} = K_{1}^{T} > 0$ is a design matrix, and $ς, τ_{1}, ς_{1}$ are design positive constants.

The adaptive laws are designed as follows

${\dot{\hat{θ}}}_{1} = γ_{1} (\frac{3 ∥ S_{1} ∥^{\frac{4}{3}}}{4 τ_{1}^{\frac{4}{3}}} ∥ z_{1} ∥^{4} - {\hat{θ}}_{1})$
20

${\dot{\hat{ζ}}}_{1} = ξ_{1} (\frac{3}{4 ς_{1}^{\frac{4}{3}}} ∥ z_{1} ∥^{4} - {\hat{ζ}}_{1})$
21

where $γ_{1} and ξ_{1}$ are positive design parameters.

Consider the Lyapunov function candidate as

$V_{1} = \frac{1}{4} {(z_{1}^{T} z_{1})}^{2} + \frac{1}{2 γ_{1}} {\tilde{θ}}_{1}^{2} + \frac{1}{2 ξ_{1}} {\tilde{ζ}}_{1}^{2}$
22

From equation (7), and combining with equations (19) to (21), we have

$\begin{array}{l} L V_{1} = z_{1}^{T} z_{1} z_{1}^{T} (F_{1} + Φ_{1}^{^{T}} S_{1} + G_{1} ω + D_{1} - {\dot{y}}_{d}) - \frac{1}{γ_{1}} {\tilde{θ}}_{1} {\dot{\hat{θ}}}_{1} - \frac{1}{ξ_{1}} {\tilde{ζ}}_{1} {\dot{\hat{ζ}}}_{1} \\ = - z_{1}^{T} z_{1} z_{1}^{T} K_{1} z_{1} - z_{1}^{T} z_{1} z_{1}^{T} \frac{3 | | G_{1} {| |}^{\frac{4}{3}}}{4 ς^{\frac{4}{3}}} z_{1} + z_{1}^{T} z_{1} z_{1}^{T} G_{1} z_{2} + z_{1}^{T} z_{1} z_{1}^{T} Φ_{1}^{^{T}} S_{1} + z_{1}^{T} z_{1} z_{1}^{T} D_{1} - {\hat{θ}}_{1} \frac{3 ∥ S_{1} ∥^{\frac{4}{3}}}{4 τ_{1}^{\frac{4}{3}}} {(z_{1}^{T} z_{1})}^{2} \\ - {\hat{ζ}}_{1} \frac{3}{4 ς_{1}^{\frac{4}{3}}} {(z_{1}^{T} z_{1})}^{2} - \frac{1}{γ_{1}} {\tilde{θ}}_{1} {\dot{\hat{θ}}}_{1} - \frac{1}{ξ_{1}} {\tilde{ζ}}_{1} {\dot{\hat{ζ}}}_{1} \\ \leq - λ_{\min} (K_{1}) {(z_{1}^{T} z_{1})}^{2} - \frac{3 | | G_{1} {| |}^{\frac{4}{3}}}{4 ς^{\frac{4}{3}}} {(z_{1}^{T} z_{1})}^{2} + z_{1}^{T} z_{1} z_{1}^{T} G_{1} z_{2} + z_{1}^{T} z_{1} z_{1}^{T} Φ_{1}^{^{T}} S_{1} + z_{1}^{T} z_{1} z_{1}^{T} D_{1} - θ_{1} \frac{3 ∥ S_{1} ∥^{\frac{4}{3}}}{4 τ_{1}^{\frac{4}{3}}} {(z_{1}^{T} z_{1})}^{2} \\ - ζ_{1} \frac{3}{4 ς_{1}^{\frac{4}{3}}} {(z_{1}^{T} z_{1})}^{2} + {\tilde{θ}}_{1} {\hat{θ}}_{1} + {\tilde{ζ}}_{1} {\hat{ζ}}_{1} \end{array}$
23

According to Lemma 2, we have

$z_{1}^{T} z_{1} z_{1}^{T} Φ_{1}^{^{T}} S_{1} (Ω) \leq ∥ z_{1} ∥^{3} ∥ Φ_{1}^{} ∥ ∥ S_{1} (Ω) ∥ \leq θ_{1} \frac{3 ∥ S_{1} ∥^{\frac{4}{3}}}{4 τ_{1}^{\frac{4}{3}}} {(z_{1}^{T} z_{1})}^{2} + \frac{θ_{1} τ_{1}^{4}}{4}$
24

$z_{1}^{T} z_{1} z_{1}^{T} G_{1} z_{2} \leq \frac{3 | | G_{1} {| |}^{\frac{4}{3}}}{4 ς^{\frac{4}{3}}} (z_{1}^{T} z_{1}) + \frac{ς_{_{}}^{4}}{4} {(z_{2}^{T} z_{2})}^{2}$
25

$z_{1}^{T} z_{1} z_{1}^{T} D_{1} \leq ζ_{1} \frac{3}{4 ς_{1}^{\frac{4}{3}}} {(z_{1}^{T} z_{1})}^{2} + \frac{ζ_{1} ς_{1}^{4}}{4}$
26

${\tilde{θ}}_{1} {\hat{θ}}_{1} = \frac{1}{2} (θ_{1}^{2} - {\tilde{θ}}_{1}^{2} - {\hat{θ}}_{1}^{2}) \leq \frac{1}{2} (θ_{1}^{2} - {\tilde{θ}}_{1}^{2})$
27

${\tilde{ζ}}_{1} {\hat{ζ}}_{1} = \frac{1}{2} (ζ_{1}^{2} - {\tilde{ζ}}_{1}^{2} - {\hat{ζ}}_{1}^{2}) \leq \frac{1}{2} (ζ_{1}^{2} - {\tilde{ζ}}_{1}^{2})$
28

Substituting equations (24) to (28) into equation (23) yields

$L V_{1} \leq - λ_{min} (K_{1}) {(z_{1}^{T} z_{1})}^{2} + \frac{ς^{4}}{4} {(z_{2}^{T} z_{2})}^{2} + \frac{θ_{1} τ_{1}^{4}}{4} + \frac{ζ_{1} ς_{1}^{4}}{4} + \frac{1}{2} (θ_{1}^{2} - {\tilde{θ}}_{1}^{2}) + \frac{1}{2} (ζ_{1}^{2} - {\tilde{ζ}}_{1}^{2})$
29

Step 2

According to $z_{2} = ω - α_{1}$ and the $I t \hat{o}$ formula, we have

$d z_{2} = (F_{2} + Δ F_{2} + G_{2} M_{c} + d_{2} + \bar{G} - {\dot{α}}_{1}) d t + G_{2} Λ Π d B$
30

Similar to step 1, the RBFNNs will be utilized to tackle the unknown uncertainty ΔF₂, and the optimal approximation can be written as

$Δ F_{2} = Φ_{2}^{^{T}} S_{2} (Ω, ω) + δ_{2}$
31

where the approximation error vector δ₂ satisfies $∥ δ_{2} ∥ \leq δ_{2}^{}$ with $δ_{2}^{} > 0$ .

Substituting equation (31) into equation (30) yields

$d z_{2} = (F_{2} + Φ_{2}^{^{T}} S_{2} + G_{2} M_{c} + D_{2} + \bar{G} - {\dot{α}}_{1}) d t + G_{2} Λ Π d B$
32

where $D_{2} = δ_{2} + d_{2}$ is compounded disturbance, and it is bounded, that is, $∥ D_{2} ∥ \leq ζ_{2}$ with $ζ_{2} > 0$ . The estimation of the unknown parameter ζ₂ is ${\hat{ζ}}_{2} .$ And the parameter estimation error is ${\tilde{ζ}}_{2} = ζ_{2} - {\hat{ζ}}_{2}$ .

The control law M_c* is designed as follows

$M_{c} = - G_{2}^{- 1} (K_{2} z_{2} + \frac{ς^{4}}{4} z_{2} + F_{2} - {\dot{υ}}_{1} + {\hat{θ}}_{2} \frac{3 ∥ S_{2} ∥^{\frac{4}{3}} z_{2}}{4 τ_{2}^{\frac{4}{3}}} + {\hat{ζ}}_{2} \frac{3 z_{2}}{4 ς_{2}^{\frac{4}{3}}} + \hat{ρ} z_{2} (\frac{3}{4 ς_{4}^{2}} | | G_{2} {| |}_{F}^{4} ψ_{1}^{4} + \frac{3}{4 ς_{3}^{\frac{4}{3}}} ψ_{2}^{\frac{4}{3}}))$
33

where $K_{2} = K_{2}^{T} > 0$ is a design matrix, and $τ_{2}, ς_{2}, ς_{3}, ς_{4}$ are design positive constants.

The parameter adaptive laws are designed as follows

${\dot{\hat{θ}}}_{2} = γ_{2} (\frac{3 ∥ S_{2} ∥^{\frac{4}{3}}}{4 τ_{2}^{\frac{4}{3}}} z_{2}^{4} - {\hat{θ}}_{2})$
34

${\dot{\hat{ζ}}}_{2} = ξ_{2} (\frac{3}{4 ς_{2}^{\frac{4}{3}}} z_{2}^{4} - {\hat{ζ}}_{2})$
35

$\dot{\hat{ρ}} = ϱ ((\frac{3}{4 ς_{4}^{2}} | | G_{2} {| |}_{F}^{4} ψ_{1}^{4} + \frac{3}{4 ς_{3}^{\frac{4}{3}}} ψ_{2}^{\frac{4}{3}}) z_{2}^{4} - \hat{ρ})$
36

where $γ_{2}, ξ_{2}, and ϱ$ are positive design parameters.

Consider a stochastic Lyapunov function as

$V_{2} = V_{1} + \frac{1}{4} {(z_{2}^{T} z_{2})}^{2} + \frac{1}{2 γ_{2}} {\tilde{θ}}_{2}^{2} + \frac{1}{2 ξ_{2}} {\tilde{ζ}}_{2}^{2} + \frac{1}{2 ϱ} {\tilde{ρ}}^{2}$
37

From equation (7), one has

$L V_{2} = L V_{1} + z_{2}^{T} z_{2} z_{2}^{T} (F_{2} + Φ_{2}^{^{T}} S_{2} + G_{2} M_{c} + D_{2} + \bar{G} - {\dot{α}}_{1}) + \frac{1}{2} t r (Π^{T} Λ^{T} G_{2}^{T} (2 z_{2} z_{2}^{T} + z_{2}^{T} z_{2} I) G_{2} Λ Π) - \frac{1}{γ_{2}} {\tilde{θ}}_{2} {\dot{\hat{θ}}}_{2} - \frac{1}{ξ_{2}} {\tilde{ζ}}_{2} {\dot{\hat{ζ}}}_{2} - \frac{1}{ϱ} \tilde{ρ} \dot{\hat{ρ}} = L V_{1} - z_{2}^{T} z_{2} z_{2}^{T} K_{2} z_{2} - z_{2}^{T} z_{2} z_{2}^{T} \frac{ς_{1}^{4}}{4} z_{2} + z_{2}^{T} z_{2} z_{2}^{T} Φ_{2}^{^{T}} S_{2} + z_{2}^{T} z_{2} z_{2}^{T} \bar{G} + z_{2}^{T} z_{2} z_{2}^{T} D_{2} + \frac{1}{2} t r (Π^{T} Λ^{T} G_{2}^{T} (2 z_{2} z_{2}^{T} + z_{2}^{T} z_{2} I) G_{2} Λ Π) - {\hat{θ}}_{2} \frac{3 ∥ S_{2} ∥^{\frac{4}{3}}}{4 τ_{2}^{\frac{4}{3}}} {(z_{2}^{T} z_{2})}^{2} - {\hat{ζ}}_{2} \frac{3}{4 ς_{2}^{\frac{4}{3}}} {(z_{2}^{T} z_{2})}^{2} - \hat{ρ} (\frac{3}{4 ς_{4}^{2}} | | G_{2} {| |}_{F}^{4} ψ_{1}^{4} + \frac{3}{4 ς_{3}^{\frac{4}{3}}} ψ_{2}^{\frac{4}{3}}) {(z_{2}^{T} z_{2})}^{2} - \frac{1}{γ_{2}} {\tilde{θ}}_{2} {\dot{\hat{θ}}}_{2} - \frac{1}{ξ_{2}} {\tilde{ζ}}_{2} {\dot{\hat{ζ}}}_{2} - \frac{1}{ϱ} \tilde{ρ} \dot{\hat{ρ}}$
38

According to Young’s inequality, one has

$z_{2}^{T} z_{2} z_{2}^{T} Φ_{2}^{^{T}} S_{2} \leq ∥ z_{2} ∥^{3} ∥ Φ_{2}^{} ∥ ∥ S_{2} ∥ \leq θ_{2} \frac{3 ∥ S_{2} ∥^{\frac{4}{3}}}{4 τ_{2}^{\frac{4}{3}}} ∥ z_{2} ∥^{4} + \frac{θ_{2} τ_{2}^{4}}{4}$
39

$z_{2}^{T} z_{2} z_{2}^{T} D_{2} \leq ζ_{2} \frac{3}{4 ς_{2}^{\frac{4}{3}}} ∥ z_{2} ∥^{4} + \frac{ζ_{2} ς_{2}^{4}}{4}$
40

$z_{2}^{T} z_{2} z_{2}^{T} \bar{G} \leq \frac{3}{4 ς_{3}^{\frac{4}{3}}} ∥ z_{2} ∥^{4} g_{2}^{}^{\frac{4}{3}} ρ_{2}^{\frac{4}{3}} ψ_{2}^{\frac{4}{3}} + \frac{ς_{3}^{4}}{4}$
41

$\begin{array}{l} \frac{1}{2} t r (Π^{T} Λ^{T} G_{2}^{T} (2 z_{2} z_{2}^{T} + z_{2}^{T} z_{2} I) G_{2} Λ Π) \leq \frac{3}{2} ∥ z_{2} ∥^{2} ∥ G_{2} Λ Π ∥_{F}^{2} \\ \leq \frac{3}{4 ς_{4}^{2}} ∥ G_{2} Λ Π ∥_{F}^{4} ∥ z_{2} ∥^{4} + \frac{3 ς_{4}^{2}}{4} \\ \leq \frac{3}{4 ς_{4}^{2}} ∥ z_{2} ∥^{4} ∥ G_{2} ∥_{F}^{4} ψ_{1}^{4} ∥ Π ∥_{F}^{4} ρ_{1}^{4} + \frac{3 ς_{4}^{2}}{4} \end{array}$
42

${\tilde{θ}}_{2} {\hat{θ}}_{2} = \frac{1}{2} (θ_{2}^{2} - {\tilde{θ}}_{2}^{2} - {\hat{θ}}_{2}^{2}) \leq \frac{1}{2} (θ_{2}^{2} - {\tilde{θ}}_{2}^{2})$
43

${\tilde{ζ}}_{2} {\hat{ζ}}_{2} = \frac{1}{2} (ζ_{2}^{2} - {\tilde{ζ}}_{2}^{2} - {\hat{ζ}}_{2}^{2}) \leq \frac{1}{2} (ζ_{2}^{2} - {\tilde{ζ}}_{2}^{2})$
44

$\tilde{ρ} \hat{ρ} \leq \frac{1}{2} (ρ^{2} - {\tilde{ρ}}^{2} - {\hat{ρ}}^{2}) \leq \frac{1}{2} (ρ^{2} - {\tilde{ρ}}^{2})$
45

Substituting equations (39) to (45) into equation (38), we have

$\begin{matrix} L V_{2} \leq L V_{1} - z_{2}^{T} z_{2} z_{2}^{T} K_{2} z_{2} - z_{2}^{T} z_{2} z_{2}^{T} \frac{ς^{4}}{4} z_{2} \\ + \frac{1}{2} (θ_{2}^{2} - {\tilde{θ}}_{2}^{2}) + \frac{1}{2} (ζ_{2}^{2} - {\tilde{ζ}}_{2}^{2}) \\ + \frac{1}{2} (ρ^{2} - {\tilde{ρ}}^{2}) + \frac{θ_{2} τ_{2}^{4}}{4} + \frac{ζ_{2} ς_{2}^{4}}{4} + \frac{ς_{3}^{4}}{4} + \frac{3 ς_{4}^{2}}{4} \end{matrix}$
46

By combining with equation (29), it yields

$\begin{array}{l} L V_{2} \leq - \sum_{j = 1}^{2} λ_{min} (K_{j}) {(z_{j}^{T} z_{j})}^{2} - \frac{1}{2} \sum_{j = 1}^{2} {\tilde{θ}}_{j}^{2} - \frac{1}{2} \sum_{j = 1}^{2} {\tilde{ζ}}_{j}^{2} \\ - \frac{1}{2} {\tilde{ρ}}^{2} + \frac{1}{2} \sum_{j = 1}^{2} θ_{j}^{2} + \frac{1}{2} \sum_{j = 1}^{2} ζ_{j}^{2} + \frac{1}{2} ρ^{2} \\ + \sum_{j = 1}^{2} \frac{θ_{j} τ_{j}^{4}}{4} + \sum_{j = 1}^{2} \frac{ζ_{j} ς_{j}^{4}}{4} + \frac{ς_{3}^{4}}{4} + \frac{3 ς_{4}^{2}}{4} \\ \leq - a_{0} V + b_{0} \end{array}$
47

where

${\begin{array}{l} a_{0} = \min {λ_{\min} (K_{1}), λ_{\min} (K_{2}), γ_{j}, ξ_{j}, ρ, j = 1,2} \\ b_{0} = \frac{1}{2} \sum_{j = 1}^{2} θ_{j}^{2} + \frac{1}{2} \sum_{j = 1}^{2} ζ_{j}^{2} + \frac{1}{2} ρ^{2} + \sum_{j = 1}^{2} \frac{θ_{j} τ_{j}^{4}}{4} + \sum_{j = 1}^{2} \frac{ζ_{j} ς_{j}^{4}}{4} + \frac{ς_{3}^{4}}{4} + \frac{3 ς_{4}^{2}}{4} \end{array}$
48

Thus, from equation (47), and according to the stochastic Lyapunov stability theory, the signals of the closed-loop system are uniformly bounded in the sense of probability. It may directly show that the tracking error z₁ is as small as possible by choosing suitable design parameters, and the estimate errors ${\tilde{θ}}_{i}, {\tilde{ζ}}_{i}, i = 1, 2$ and $\tilde{ρ}$ are all bounded.

Simulation study

In this section, simulation results are presented to illustrate the effectiveness of the proposed adaptive neural attitude control scheme for NSV with stochastic disturbances. The attitude dynamics of NSV with stochastic input torque is in the form of system (5).

Suppose that the aerodynamic coefficients and aerodynamic moment coefficients have +20% and −20% uncertainties. On the other hand, suppose that all the external disturbances are assumed to act on the fast-loop system in the form of the moment disturbances d₂ which are given as

$d_{2} = [5 \times 10^{5} (sin (5 t) + 0.3),5 \times 10^{5} (cos (5 t) + 0.1), 5 \times 10^{5} sin (6 t)] Nm$
49

The stochastic input disturbance is in the form of the Gaussian white noise $Λ (Ω, ω) ξ \in ℝ^{3}$ where the PSD of the white noise ξ is $Π = (Π_{i j})_{3 \times 3}$ with $Π_{i i} = 0.1$ and $Π_{i j} = 0, (i \neq j)$ , and the coefficient matrix is chosen as⁵⁷

$Λ (Ω, ω) = [\begin{matrix} p + q & 0 & p \\ 0 & r & 0 \\ q sin (r) & p & 0 \end{matrix}]$
50

Thus, we obtain

$\bar{G} = G_{2} (Ω, ω) {[p + 0.5 q,0.5 p,0.5 q^{2} sin (r) \cos (r)]}^{T}$
51

It is clear that the following inequalities hold

$∥ Λ (Ω, ω) ∥_{F} = \sqrt{| p + q |^{2} + q^{2} sin {(r)}^{2} + 2 p^{2} + r^{2}} \leq \sqrt{4 p^{2} + 3 q^{2} + r^{2}} \leq 2 \sqrt{p^{2} + q^{2} + r^{2}} = ρ_{1} ψ_{1}$
52

$| | \bar{G} | | \leq 0.5 g_{2}^{} \sqrt{7 p^{2} + 3 q^{2} + q^{4}} = g_{2}^{} ρ_{2} ψ_{2}$
53

where $ρ_{1} = 2, ρ_{2} =0.5, ψ_{1} = \sqrt{p^{2} + q^{2} + r^{2}}, and ψ_{2} = \sqrt{7 p^{2} + 3 q^{2} + q^{4}}$ .

The initial values are chosen as H ₀ = 21,000 m, V ₀ = 4000 m/s, α ₀ = 2°, β ₀ = −1°, μ ₀ = 3°, p = q = r = 0°/s. The desired flight attitudes Ω_d are chosen as

$Ω_{d} = {\begin{array}{l} α_{d} = {\begin{matrix} - 2^{^{\circ}}, & 6 k \leq t \leq 6 k + 3 \\ 2^{^{\circ}}, & 6 k + 3 \leq t \leq 6 k + 6 \end{matrix} \\ β_{d} {= 0}^{^{\circ}} \\ μ_{d} = 1.7 sin (0.8 π t) + sin (0.3 π t) \end{array}$
54

The designed parameters and matrices of the controller are chosen as $K_{1} = K_{2} = 30 diag {1, 1, 1},$ $γ_{1} = γ_{2} = 1,$ $ξ_{1} = ξ_{2} = 1, ρ = 2,$ ς = 1, $τ_{1} = τ_{2} = ς_{1} = ς_{2} = 5,$ and $ς_{3} = ς_{4} = 10$ . The intermediate control function υ₁, the controller M_c*, and the adaptive laws are designed as equations (19), (33), (20), (21), (34), (35), and (36), respectively. From Figures 1 to 3, we observe that the tracking performance has been achieved under the effects of the stochastic disturbances and the system uncertainties. In addition, Figure 4 shows the variation of actual control input signals. From above simulation results, the proposed adaptive neural tracking control scheme is valid for stochastic nonlinear MIMO systems with system uncertainties and stochastic disturbances.

Figure 1.
Response curves of attitude angles under adaptive neural controller.

Figure 2.
Response curves of tracking errors under adaptive neural controller.

Figure 3.
Response curves of attitude angles rate under adaptive neural controller.

Figure 4.
Actual control inputs under adaptive neural controller.

Conclusions

In this article, an adaptive neural controller has been proposed for the NSV with stochastic disturbances and unknown uncertainties. The packaged unknown uncertainties are tackled using the RBFNNs. Furthermore, an effective robust adaptive backstepping-based control scheme has been designed. On the basis of the stochastic Lyapunov stability theory, the stability of the closed-loop system is proved, and the tracking error can converge to a small neighborhood of the origin. Finally, the effectiveness of the proposed robust attitude control scheme was illustrated via the simulation results.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported in part by the National Natural Science Foundation of China under grants 61573184 and 61374212, in part by the Fundamental Research Funds for the Central Universities under grant NE2016101, and in part by the Projects of Anhui Province University Outstanding Youth Talent Support Program under grant gxyq2017068.

References

Shi

. An overview on flight dynamics and control approaches for hypersonic vehicles. Sci China Inform Sci 2015; 58(7): 1–19.

Sun

Yang

. Adaptive discrete-time controller design with neural network for hypersonic flight vehicle via backstepping. Int J Control 2011; 84(9): 1543–1552.

Chen

Jiang

. Robust attitude control of near space vehicles with time-varying disturbances. Int J Control Automat Syst 2013; 11: 182–187.

Zhao

Jiang

Chowdhury

. Active fault-tolerant control for near space vehicles based on reference model adaptive sliding mode scheme. Int J Adapt Control Signal Process 2014; 28: 765–777.

Jiang

Tao

. Fault tolerant control for a class of nonlinear systems with application to near space vehicle. Circuits Syst Signal Process 2011; 30: 655–672.

Sun

Liu

. Adaptive Kriging controller design for hypersonic flight vehicle via backstepping. IET Control Theory Appl 2012; 6: 487–497.

Zhang

Pan

. Neural network based dynamic surface control of hypersonic flight dynamics using small-gain theorem. Neurocomputing 2016; 173: 690–699.

. Robust adaptive neural control of flexible hypersonic flight vehicle with dead-zone input nonlinearity. Nonlinear Dynam 2015; 80(3): 1509–1520.

Wang

. Adaptive fuzzy tracking control for a constrained flexible air-breathing hypersonic vehicle based on actuator compensation. Int J Adv Robot Syst 2016; 13(6): 1–10.

10.

Gao

Jiang

Shi

. Passive fault-tolerant control design for near-space hypersonic vehicle dynamical system. Circuits Syst Signal Process 2012; 31: 565–581.

11.

Jiang

. Adaptive functional link network control of near-space vehicles with dynamical uncertainties. J Syst Eng Electron 2010; 21: 868–876.

12.

Yang

Chen

. Adaptive neural prescribed performance tracking control for near space vehicles with input nonlinearity. Neurocomputing 2016; 174: 780–789.

13.

Pan

Wang

. Discrete-time hypersonic flight control based on extreme learning machine. Neurocomputing 2014; 128: 232–241.

14.

Jiang

Gao

Shi

. Adaptive fault-tolerant tracking control of near-space vehicle using Takagi-Sugeno fuzzy models. IEEE Trans Fuzzy Syst 2010; 18: 1000–1007.

15.

Gao

Sun

. Fuzzy tracking control design for hypersonic vehicles via T-S model. Sci China Inform Sci 2011; 54: 521–528.

16.

Yang

Pan

. Global neural dynamic surface tracking control of strict-feedback systems with application to hypersonic flight vehicle. IEEE Trans Neural Networks Learn Syst 2015; 26(10): 2563–2575.

17.

Chen

Jiang

. Guaranteed transient performance based control with input saturation for near space vehicles. Sci China Inform Sci; 57: 1–12.

18.

Chen

Mei

Jiang

. Sliding mode control for a class of uncertain MIMO nonlinear systems with application to near-space vehicles. Math Probl Eng 2013; 2013: 180589.

19.

Cheng

Jiang

. Online-SVR-compensated nonlinear generalized predictive control for hypersonic vehicles. Sci China Inform Sci 2011; 54: 551–562.

20.

Shi

Yang

. Composite fuzzy control of a class of uncertain nonlinear systems with disturbance observer. Nonlinear Dynam 2015; 80: 341–351.

21.

Yang

. Sliding-mode control for systems with mismatched uncertainties via a disturbance observer. IEEE Trans Ind Electron 2013; 60: 160–169.

22.

Yang

Sun

. Nonlinear-disturbance-observer-based robust flight control for airbreathing hypersonic vehicles. IEEE Trans Aerosp Electron Syst 2013; 49: 1263–1275.

23.

Yang

Shen

. Disturbance observer-based L1 robust tracking control for hypersonic vehicles with T-S disturbance modeling. Int J Adv Robot Syst 2016; 13(6): 1–10.

24.

Guo

Chen

. Disturbance attenuation and rejection for systems with nonlinearity via DOBC approach. Int J Robust Nonlinear Control 2005; 15: 109–125.

25.

Yang

. Robust autopilot design for bank-to-turn missiles using disturbance observers. IEEE Trans Aerosp Electron Syst 2013; 49: 558–579.

26.

Shao

Chen

Yan

. Adaptive sliding mode synchronization for a class of fractional-order chaotic systems with disturbance. Nonlinear Dynam 2016; 83: 1855–1866.

27.

Chen

. Adaptive neural output feedback control of uncertain nonlinear systems with unknown hysteresis using disturbance observer. IEEE Trans Ind Electron 2015; 62(12): 7706–7716.

28.

Chen

. Adaptive fuzzy tracking control for a class of uncertain MIMO nonlinear systems using disturbance observer. Sci China Inform Sci 2014; 57(1): 1–13.

29.

Shao

Chen

. Adaptive neural control for an uncertain fractional-order rotational mechanical system using disturbance observer. IET Control Theory Appl 2016; 10(16): 1972–1980.

30.

Liu

Zhang

Yao

. Composite anti-disturbance control for a class of nonlinear stochastic systems via disturbance observer. Math Probl Eng 2013; 2013: 281340.

31.

Chen

. Adaptive dynamic surface control of NSVs with input saturation using a disturbance observer. Chinese J Aeronaut 2015; 28: 853–864.

32.

Chen

Ren

. Anti-disturbance control of hypersonic flight vehicles with input saturation using disturbance observer. Sci China Inform Sci 2015; 58(7): 1–12.

33.

Cheng

Jiang

. The research of SMDO based NGPC method for NSV control system. J Astronaut 2010; 31: 423–431.

34.

Tong

. Adaptive fuzzy output feedback tracking backstepping control of strict-feedback nonlinear systems with unknown dead zones. IEEE Trans Fuzzy Syst 2012; 20: 168–180.

35.

Liu

Tong

. Adaptive fuzzy control via observer design for uncertain nonlinear systems with unmodeled dynamics. IEEE Trans Fuzzy Syst 2013; 21: 275–288.

36.

Liu

Tong

Chen

. Neural controller design-based adaptive control for nonlinear MIMO systems with unknown hysteresis inputs. IEEE Trans Cybernet 2016; 46(1): 9–19.

37.

Liu

Tong

. Neural network control-based adaptive learning design for nonlinear systems with full-state constraints. IEEE Trans Neural Networks Learn Syst 2016; 27(7): 1562–1571.

38.

Liu

Tong

. Optimal control-based adaptive NN design for a class of nonlinear discrete-time block-triangular systems. IEEE Trans Cybernet 2016; 46(11): 2670–2680.

39.

Liu

Tong

. Barrier Lyapunov functions for Nussbaum gain adaptive control of full state constrained nonlinear systems. Automatica 2017; 76: 143–152.

40.

Shao

Chen

Yan

. Prescribed performance synchronization for uncertain chaotic systems with input saturation based on neural networks. Neural Comput Appl 2016; 24(9). DOI: 10.1007/s00521-016-2629-5.

41.

Chen

Liu

. Approximation-based adaptive neural control design for a class of nonlinear systems. IEEE Trans Cybernet 2014; 44: 610–679.

42.

Deng

Krstic

Williams

. Stabilization of stochastic nonlinear systems driven by noise of unknown covariance. IEEE Trans Automat Control 2001; 46(8): 1237–1253.

43.

Tong

. Observer-based adaptive fuzzy backstepping control for a class of stochastic nonlinear strict-feedback systems. IEEE Trans Syst Man Cybernet B 2011; 41(6): 1693–1704.

44.

Chen

Liu

Wen

. Fuzzy neural network-based adaptive control for a class of uncertain nonlinear stochastic systems. IEEE Trans Cybernet 2014; 44(5): 583–593.

45.

Psillakis

Alexandridis

. Adaptive neural tracking for a class of SISO uncertain and stochastic non-linear systems. In: Proceedings of 44 IEEE conference on decision and control, Seville, Spain, 12–15 December 2005. pp. 7822–7827.

46.

Tong

Jing

. Adaptive fuzzy control of uncertain stochastic nonlinear systems with unknown dead zone using small-gain approach. Fuzzy Sets Syst 2013; 235: 1–24.

47.

Chen

. Adaptive NN output-feedback decentralized stabilization for a class of large-scale stochastic nonlinear strict-feedback systems. Int J Robust Nonlinear Control 2011; 21(4): 452–472.

48.

Psillakis

Alexandridis

. NN-based adaptive tracking control of uncertain nonlinear systems disturbed by unknown covariance noise. IEEE Trans Neural Networks 2007; 18(6): 1830–1835.

49.

Yang

Shi

. Adaptive tracking for stochastic nonlinear systems with Markovian switching. IEEE Trans Automat Control 2010; 55(9): 2135–2141.

50.

Wang

Chen

Liu

. Robust adaptive fuzzy tracking control for pure-feedback stochastic nonlinear systems with input constraints. IEEE Trans Cybernet 2013; 43(6): 2093–2104.

51.

Fan

Han

Wen

. Decentralized adaptive output-feedback controller design for stochastic nonlinear interconnected systems. Automatica 2012; 48(11): 2866–2873.

52.

Cui

Wang

Zhuang

. Adaptive decentralized NN control of large-scale stochastic nonlinear time-delay systems with unknown dead-zone inputs. Neurocomputing 2015; 158: 194–203.

53.

Tong

. Observer-based adaptive fuzzy backstepping dynamic surface control design and stability analysis for MIMO stochastic nonlinear systems. Nonlinear Dynam 2012; 69(3): 1333–1349.

54.

Yan

Chen

. Dynamic surface control for a class of stochastic non-linear systems with input saturation. IET Control Theory Appl 2016; 10(1): 35–43.

55.

Shahvali

Askari

. Adaptive neural dynamic surface control of MIMO stochastic nonlinear systems with unknown control directions. Int J Adapt Control Signal Process 2017; 31(1): 97–121.

56.

Cui

Xie

. Modeling and adaptive tracking for a class of stochastic Lagrangian control systems. Automatica 2013; 49(3): 770–779.

57.

Samiei

Izadi

Viswanathan

. Robust stabilization of rigid body attitude motion in the presence of a stochastic input torque. In: IEEE international conference on robotics and automation, 26–30 May 2015, pp. 428–433, Seattle, Washington: Washington State Convention Center.

58.

. Nonlinear robust stochastic control for unmanned aerial vehicles. J Guid Control Dynam 2009; 32(4): 1308–1319.

59.

Zhu

. Nonlinear stochastic dynamics and control: framework of Hamiltonian theory. Beijing: Science Press, 2003.

60.

Khasminskii

. Stochastic stability of differential equations. Berlin: Springer, 2011.

61.

Mao

Stochastic differential equations and their applications. Chichester: Horwood, 2007.

Adaptive neural tracking control for near-space vehicles with stochastic disturbances

Abstract

Keywords

Introduction

Problem statements and preliminaries

Problem formulation and some assumptions

Assumption 1

Assumption 2

Assumption 3

Assumption 4

Mathematical preliminaries

Definition 1

Definition 2

Definition 3

Lemma 1

Lemma 2 (Young’s inequality)

Neural networks

Adaptive neural tracking control design

Step 1

Step 2

Simulation study

Conclusions

Footnotes

Declaration of conflicting interests

Funding

References