Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Adaptive Backstepping Control for Fractional-Order Nonlinear Systems with External Disturbance and Uncertain Parameters Using Smooth Control

Adaptive Backstepping Control for Fractional-Order Nonlinear Systems with External Disturbance... Adaptive Backstepping Control for Fractional-Order Nonlinear Systems with External Disturbance and Uncertain Parameters Using Smooth Control Xinyao Li Changyun Wen School of Electrical and Electronic Engineering School of Electrical and Electronic Engineering Nanyang Technological University Nanyang Technological University Singapore, 639798 Singapore, 639798 E180209@e.ntu.edu.sg ecywen@ntu.edu.sg Ying Zou School of Electrical and Electronic Engineering Nanyang Technological University Singapore, 639798 zouy0011@e.ntu.edu.sg April 15, 2020 ABSTRACT In this paper, we consider controlling a class of single-input-single-output (SISO) commensurate fractional-order nonlinear systems with parametric uncertainty and external disturbance. Based on backstepping approach, an adaptive controller is proposed with adaptive laws that are used to estimate the unknown system parameters and the bound of unknown disturbance. Instead of using discontinuous functions such as the sign function, an auxiliary function is employed to obtain a smooth control input that is still able to achieve perfect tracking in the presence of bounded disturbances. Indeed, global boundedness of all closed-loop signals and asymptotic perfect tracking of fractional-order system output to a given reference trajectory are proved by using fractional directed Lyapunov method. To verify the effectiveness of the proposed control method, simulation examples are presented. Keywords Adaptive backstepping control; fractional-order; nonlinear systems; smooth control. 1 Introduction Fractional calculus owns a history of more than 300 years. It is a branch of mathematics that deals with non-integer order derivatives and integrals. Compared with integer order calculus, fractional order integral and derivative can both be treated as weighted integral and thus they have the properties of hereditary and infinite memory [1, 2], which can also be seen from their definitions given in Section 2. Such properties give the most significant meaning of fractional-order derivative compared to integer-order derivative that does not have such properties. Thus, using fractional order models could better and more accurately describe the characteristics of the real world systems than integer order models, as elaborated in [3–5]. In the past few years, many researchers have paid significant attention to fractional order calculus and constructed models for real-world systems, including viscoelasticity, complex systems, neural networks, transmission line, multi agent systems, and so on, see for examples [6–10]. Moreover, fractional order systems as well as their controls have also been studied in recent years [11–13]. However, it is difficult to simply apply the approaches of controller design and analysis developed for integer-order systems to fractional-order systems due to the lack of appropriate mathematical tools. For example, the fractional-order derivative of a composite function is the sum of infinite number of terms, which is different from the concise closed form expression of its integer-order derivative easily obtained by applying the chain rule. Backstepping technique, which demonstrates a step-by-step design procedure by constructing Lyapunov function and virtual control signal at each step, is widely applied in controlling integer-order nonlinear systems, see [14] and [15] for arXiv:2002.00162v3 [eess.SY] 14 Apr 2020 example. However, it is not easy to directly employ backstepping control technique on fractional-order systems due to the challenge of obtaining the fractional derivative of the quadratic-type Lyapunov function. In [15], the backstepping technique is extended to fractional order systems without entirely considering uncertainties. Integer-order Lyapunov ( ) method is applied in [15] by proving zz _ < 0 if zz < 0, where z denotes virtual error defined in backstepping algorithm and 0 < < 1. A method that transforms backstepping control problem for the fractional-order systems to integer-order by taking into account the frequency distributed model is proposed in [16–18]. In this way, the stability analysis for fractional-order systems is also carried out with integer-order Lyapunov method. In order to handle disturbances in fractional-order nonlinear systems, nonlinear disturbances observers are designed to counteract the effects caused by unknown disturbances in [19] and [20]. However, strong assumptions that the disturbances must be constant or they should have bounded derivatives should be met to achieve satisfactory perfor- mances. Adaptive controllers utilizing sign function are developed in [21–24], which ensure the stabilization/tracking error converges to zero with non-continuous control because of the non-continuity of sign function. For the purpose of avoiding calculating the fractional-order derivatives of virtual control signals, the involved derivatives are directly subtracted in the design of virtual controllers in subsequent steps in [25–27]. However, the boundedness of signals in the closed-loop control systems is not theoretically shown in these works. Fuzzy logic is employed in [21] and [28] to deal with system uncertainties as well as computing fractional order derivatives of virtual control signals with the assumption that the errors resulted from fuzzy approximation of the true values of system uncertainties and the fractional order derivatives of virtual control signals are bounded. In addition, both works can only show that the output tracking errors tend to an arbitrary small region. Besides, [28] combines dynamic surface control with adaptive backstepping method for SISO fractional-order nonlinear systems with uncertain system functions and multiple external unknown disturbances. Although it can eliminate the chattering phenomenon in control signals by getting rid of using the sign function, their results can only be obtained if the initial conditions are located in certain range. Dynamic surface control integrated with neuro-fuzzy network for fractional-order nonlinear system subjects to input constraint is addressed in [29]. Similarly, the output tracking error is only guaranteed to converge to a small region around the origin with constraint of the initial conditions. Therefore, it is still open and challenging to construct an adaptive backstepping controller with smooth control signal for fractional nonlinear systems involving both parametric uncertainties and external time-varying disturbance, which ensures asymptotic stability and perfect tracking property without restrictions on the initial conditions. Inspired by the discussions above, in this paper we address such an issue. To have a smooth control signal, an auxiliary function is used in lieu of the sign function in the designed controller. We employ the fractional directed Lyapunov method for our design and analysis. As we know, when backstepping approach is employed to design controller for high-order nonlinear systems, the time derivatives of virtual control signals are required and they can easily be obtained for integer-order systems by chain rule. But, as mentioned above, this is not the case for fractional-order systems. To overcome this difficulty, a novel approach for approximating the fractional-order time derivatives of virtual control signals is proposed and adaptive laws are designed to estimate the bounds of approximation errors. It is shown that, with the designed controller and adaptive laws, the closed loop system is globally stable and its output tracks a given reference input asymptotically, even in the presence of external time-varying disturbance and uncertain parameters. To the best of the authors’ knowledge, this is the first paper to have such results. Simulation studies illustrate the effectiveness of the proposed control scheme and also reveal its advantages compared to an existing approach. In summary, the main contribution of this paper is to design a smooth control that achieves asymptotic perfect output tracking/stabilization for fractional-order nonlinear commensurate systems and ensures global stability in the sense that all signals in the closed-loop systems remain globally bounded, even in the presence of unknown bounded external disturbances and also system uncertainties. The rest of the paper is organized as follows. Preliminaries and the fractional-order system description are provided in Section 2. In Section 3, the design of an adaptive controller is presented in detail. In Section 4, the scheme is illustrated by simulation studies with comparison to that in [21]. Finally the paper is concluded in Section 5. 2 Preliminaries and Problem Formulation 2.1 Preliminaries Definition 1 [30]: The fractional integral of an integrable function f (t) with 2 R is 1 f ( ) I f (t) = d (1) 0 t ( ) (t  ) where I means the fractional integral of order with initial time t and () denotes the well-known Gamma t 0 0 t t z1 function, which is defined as (z) = e t dt, where z 2 C. One of the significant properties of Gamma 2 function is [31]: (z + 1) = z(z); (n + 1) = n(n) = n(n 1)(n 1) =  = n!; (n) = 1, where n 2 N = fnjn  0; n 2 Ng. Definition 2 [30]: The Caputo fractional derivative of a function is shown as m (m) d 1 f ( ) (m ) D f (t) = I f (t) = d (2) t t 0 t 0 m m+1 dt (m ) (t  ) where m 1 < < m 2 Z . From equation (2) we can observe that the Caputo derivative of a constant is 0. Another commonly used fractional derivative is named Riemann-Liouville (RL) and the RL fractional derivative of a function RL f (t) is denoted as D f (t). Different from the Caputo derivative, RL derivative of a constant is not equal to 0 [30, 32]. t t To obtain the unique solution for fractional differential equation D x(t) = f (x; t), (m 1 < < m 2 0 t Z and t  t ), the initial values need to be determined. According to [30, 33] and [34], fractional differential equations with Caputo-type derivative have initial values that are in-line with integer-order differential equations, i.e. 0 (m1) x(t ); x (t ); : : : ; x (t ), which contain specific physical interpretations. On the contrary, although mathematically 0 0 0 the initial value problem for RL fractional differential equations is rigorous and solvable, it lacks of practical explanation since the physical meanings of these initial conditions are unknown yet. Therefore, Caputo-type fractional systems are frequently employed in practical analysis. n + Lemma 1 [35]: Assume x(t) 2 R is a smooth function, n 2 Z , then for 8t  t , C T T C D [x (t)Px(t)]  x (t)P D x(t); 0 < < 1 (3) t t t t 0 0 nn where P 2 R is a positive definite constant matrix. Lemma 2 [30]: If 0 < < 2, represents an arbitrary complex number and real number  satisfies that <  < minf;  g, then for any integer n  1, 1n E (z) = + o(jzj ); (4) ( j) j=1 where   jarg(z)j   and jzj ! 1. Here E (z) is the Mittag-Leffler function, which is described as follows: E (z) = ; ( > 0; > 0): (5) ( k + ) k=0 Lemma 3 [33, 35]: If the Caputo fractional derivative D f (t) is integrable, then t t m1 (k) f (t ) C k I D f (t) = f (t) (t t ) (6) t 0 0 t t t k! k=0 where m 1 < < m 2 Z . Particularly, for 0 <  1, we can obtain I D f (t) = f (t) f (t ): (7) t 0 0 t t t Definition 3 [36–38]: For fractional nonautonomous system D x (t) = f (x; t), where i = 1; 2; : : : ; n, 0 < < 1, t i i 0 t T n initial condition is x(t ) = [x (t ); x (t ); : : : ; x (t )] 2 R , D indicates Caputo or RL fractional derivative, 0 1 0 2 0 n 0 t 0 t f (x; t) : [t ;1) ! R is locally Lipschitz in x and piecewise continuous in t (which insinuates the existence i 0 and uniqueness of the solution to the fractional systems [30]) and 2 R stands for a region that contains the origin T     T x = [0; 0; : : : ; 0] . The equilibrium x = [x ; x ; : : : ; x ] of this system is defined as D x = f (x ; t) for t  t . t i 0 1 2 n 0 t Without loss of generality, we set the equilibrium x = [0; 0; : : : ; 0] . + + Lemma 4 [39]: Assume f (t) : R ! R is uniformly continuous and lim I f (t) = 0 with 0 < < 1 and t!1 t 0 t 8t > t > 0, then lim f (t) = 0. 0 t!1 n n Lemma 5 [40]: If x(t) : [t ;1)  R ! R is uniformly continuous, I jx(t)j  K for 8t > t > 0 and 0 t 0 0 t 0 < < 1 with constants p; K > 0, then lim x(t) = 0. t!1 Lemma 6 [41]: The following inequality holds for x 2 R: jxj <  + xsg(x; ) (8) 3 where sg(x; ) = p (9) 2 2 x + and  is a differentiable function which meets  > 0 and ( )d < 1; 8t  0. From equation (9), it can be derived that sg(x; ) is differentiable and thus it is a smooth function for8t  0. 2.2 Problem Formulation In this paper, Caputo-type definition of the fractional derivatives is utilized. Consider the following class of n-th order SISO uncertain commensurate fractional-order nonlinear systems with time-varying disturbance in strict feedback form: C T D x (t) = x (t) +  (x (t)) > 1 2 1 1 0 t 1 C T D x (t) = x (t) +  (x (t)) > 2 3 2 2 0 t 2 < . (10) C T D x (t) = x (t) +  (x (t)) n1 n n1 n1 0 t n1 C T D x (t) = g(x(t))u(t) +  (x(t)) + d(t) n n 0 t n y(t) = x (t) where the fractional-orders of all the state equations are equal to 2 (0; 1], u(t) 2 R and y(t) 2 R represent the control T n input and system output respectively, x(t) = [x (t); x (t); ; x (t)] 2 R denotes the measurable state vector, 1 2 n T i p x (t) = [x (t); x (t); ; x (t)] 2 R ,  (x (t)) 2 R (i = 1; 2; ; n) is a vector with its elements being known i 1 2 i i i nonlinear smooth functions,  2 R (i = 1; 2; ; n) is an unknown constant vector, g(x(t)) is a known non-zero smooth nonlinear function and d(t) stands for an unknown bounded time-varying external disturbance with unknown bound D. Remark 1: Different from the control strategies in [42–45] that are designed with disturbance observer or disturbance resilience, a fractional order adaptive law, which can be found in (31), is to be derived to estimate the unknown upper bound D of the disturbance. The external time-varying disturbance d(t) is allowed to be norm-bounded without any condition on its derivative in this paper. By employing the estimated bound in designing control input u, such unknown disturbance can be compensated, which helps us to solve the following control problem. The control problem is to design an adaptive controller for the class of systems described in (10) such that the following objectives are achieved: 1) the closed-loop system is globally stable in the sense that all the signals including parameter estimates are bounded; 2) the system output y(t) asymptotically tracks a reference signal r(t), i.e. lim [y(t) r(t)] = 0. t!1 Assumption 1: The given reference signal r(t) and its -th order Caputo-type fractional derivative D r(t) are 0 t smooth and bounded. 3 Adaptive Controller Design and Stability Analysis 3.1 Adaptive Controller Design To achieve the above objectives, an adaptive controller is designed based on the backstepping design procedure. Defining virtual errors as follows: z = x r, which is also known as tracking error, z = x (i = 2; : : : ; n), 1 1 i i i1 where (j = 1; 2; : : : ; n 1) denotes the virtual control signal that will be designed iteratively later. 1 2 Step 1: Consider the first Lyapunov function V = z . According to Lemma 1, the Caputo derivative of V is 1 1 1 C C 2 C C C T C D V = D z  z D z = z [ D x D r] = z [x +  (x ) D r] 1 1 1 1 1 1 1 2 1 1 0 t 0 t 0 t 0 t 0 t 1 0 t (11) T C T C = z [z + +  (x ) D r] = z z + z + z  (x ) z D r: 1 2 1 1 1 1 2 1 1 1 1 1 1 1 0 t 1 0 t Let virtual control signal be T C = c z  (x ) + D r; (12) 1 1 1 1 1 1 0 t 4 ^ ~ ^ where  stands for the estimates of  and c is a positive design parameter. Defining  =   , then we have 1 1 1 1 1 1 C C ~ ^ D  = D  . Therefore, inequality (11) becomes 1 1 0 t 0 t C T C 2 T C T C D V z z + z + z  (x ) z D r = z z c z z  (x ) + z D r + z  (x ) z D r 1 1 2 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 0 t 1 0 t 1 0 t 1 0 t 2 T =z z c z + z  (x ) : 1 2 1 1 1 1 1 (13) 1 T 1 ~ ~ Let V = V +   , where is positive definite matrix, then the Caputo derivative of V is 1 1 1 1 1 2 1 C C C T 1 C T 1C 2 T T 1C ~ ~ ~ ~ ~ ~ ~ D V = D V + D    D V +  D   z z c z + z  (x ) +  D 1 1 1 1 1 1 2 1 1 1 1 1 1 0 t 0 t 0 t 1 1 0 t 1 1 0 t 1 1 1 0 t 2 T T C ~ ~ ^ =z z c z + z  (x )  D  : 1 2 1 1 1 1 1 1 1 1 1 0 t (14) Designing adaptive law as D  = z  (x ); (15) 1 1 1 1 1 0 t then inequality (14) becomes C 2 T T 1 2 ~ ~ (16) D V z z c z + z  (x )   (x )z = z z c z : 1 1 2 1 1 1 1 1 1 1 1 1 1 2 1 1 0 t 1 1 Step i (i = 2; ; n 1): The fractional order derivative of z is C C C T C T C D z = D x D = x +  (x (t)) D = z + +  (x (t)) D : (17) i i i1 i+1 i i i1 i+1 i i i i1 0 t 0 t 0 t i 0 t i 0 t Inspired by the idea in [46] and [47] where D f [u(x)] is approximated by f (u) D u(x) with approxima- 0 x u 0 x i1 @ C i1 C tion error (t) which is bounded by an unknown bound  , D is approximated by ( ) D x + i1 j 0 t 0 t j=1 @x i1 @ i1 C ( ) D  . Therefore by defining  (t) as the approximation error with bound   , (17) can be written v;j i1 i1 0 t j=1 ^ v;j as C T D z = z + +   +  +  : (18) i i+1 i v;i i i1 0 t v;i The functions and parameters in (18) are given below @ @ @ i1 i1 i1 T T T T T =[ (x ); ( ) (x ); ; ( ) (x ); ( ) (x )] ; v;i i i1 2 1 i i1 2 1 @x @x @x i1 2 1 T T T T =[ ;  ; ;  ] ; v;i i i1 1 (19) i i1 X X @ @ i1 i1 = ( )x + ( ) D  ; i j v;j 0 t @x j1 @ v;j j=2 j=1 where  is the estimate of  ; j = 1; 2; ; i 1. v;j v;j Choose virtual control signal as = z c z     ^ ; (20) i i1 i i v;i i i1 v;i where c denotes a positive design parameter,  means the estimate of  and  ^ is the estimation of   with i v;i v;i i1 i1 their adaptive law respectively designed as D  = z  ; v;i i i v;i 0 t (21) D  ^ = z  ; i1 i i1 0 t 1 2 1 T 1 2 ~ ~ ~ ^ where is positive definite and  > 0. Let V = V + z +   +  ~ , where  = i i1 i i1 i v;i v;i v;i v;i v;i i i1 2 2 2 i1 and  ~ =    ^ , then i1 i1 i1 C C C T 1C C ~ ~ D V  D V + z D z +  D  +  ~ D  ~ 0 t i 0 t i1 i0 t i v;i 0 t v;i i10 t i1 i1 (22) C C T 1C C ~ ^ = D V + z D z  D   ~ D  ^ : i1 i i v;i i1 i1 0 t 0 t v;i i 0 t 0 t i1 5 P i1 C 2 Since D V  z z c z , thus we get i1 i1 i j j 0 t j=1 C 2 D V  z z c z : (23) i i i+1 j j 0 t j=1 Step n: The fractional order derivative of z is C T D z = g(x)u + d +   +  +  ; (24) n v;n n n1 0 t v;n where @ @ @ n1 n1 n1 T T T T T =[ (x); ( ) (x ); ; ( ) (x ); ( ) (x )] ; v;n n1 2 1 n n1 2 1 @x @x @x n1 2 1 T T T T =[ ;  ; ;  ] ; v;n n n1 1 n n1 X X @ @ n1 n1 = ( )x + ( ) D  ; n j v;j 0 t @x j1 @ v;j j=2 j=1 (25) n1 n1 X X @ @ n1 n1 C C = ( ) D x + ( ) D n1 j v;j 0 t 0 t @x j @ v;j j=1 j=1 D ; n1 0 t j j : n1 n1 1 2 Considering the Lyapunov function V = V + z , then the Caputo derivative of V is n n1 n n C C C 2 C C D V = D V + D z  D V + z D z n n1 n n1 n n 0 t 0 t 0 t 0 t 0 t n1 2 T z z c z + z [g(x)u + d +   +  +   ] n1 n j j n v;n n n1 v;n (26) j=1 n1 2 T z z c z + z [g(x)u +   +  +   ] +jz j D; n1 n j j n v;n n n1 n v;n j=1 where D is an unknown bound of d(t). Considering Lemma 6, since jz j <  + z sg(z ; ), hence jz j D n n n n D + z sg(z ; )D. Therefore, (26) can be written as n n n1 C 2 T D V z z c z + z [g(x)u +   +  +   ] + D + z sg(z ; )D: (27) n n1 n j j n v;n n n1 n n 0 t v;n j=1 1 1 1 1 2 T 1 2 2 ~ ~ ~ Let V = V + z +   +  ~ + D , where is positive definite,  > 0,  > 0, n n1 n v;n n n1 v;n n n1 2 2 2 2 n1 ~ ^ ~ ^ ^ ^ =   ,  ~ =    ^ , D = D D and  ,  ^ and D represent the estimates of  , v;n v;n v;n n1 n1 n1 v;n n1 v;n n1 and D respectively. Then 1 1 C C C T 1C C C ~ ~ ~ ~ D V  D V + z D z +  D  +  ~ D  ~ + D D D n n1 n n v;n n1 n1 0 t 0 t 0 t v;n n 0 t 0 t 0 t n1 (28) 1 1 C C T 1C C C ~ ^ ~ ^ = D V + z D z  D   ~ D  ^ D D D: 0 t n1 n0 t n v;n n 0 t v;n n10 t n1 0 t n1 According to (26) and (27), we then have n1 C 2 T D V z z c z + z [g(x)u +   +  +   ] + D + z sg(z ; )D n n1 n j j n v;n n n1 n n 0 t v;n j=1 (29) 1 1 T 1C C C ~ ^ ~ ^ D   ~ D  ^ D D D: v;n n1 n1 v;n n 0 t 0 t 0 t n1 Finally the real control input u is designed as ^ ^ u = [z c z   sg(z ; )D   ^ ]; (30) n1 n n v;n n n n1 v;n g(x) 6 at where function (t) = e with a being a positive constant and c is positive design parameter. Also we have the following adaptive laws designed D  = z  ; v;n n n v;n 0 t D  ^ = z  ; (31) n1 n n1 0 t D D = z sg(z ; ): n n 0 t Hence by substituting (30) and (31) into (29), we have C 2 D V  c z + D: (32) n j j 0 t j=1 Remark 2: Function sg(z ; ) in the control law (30) is engaged to compensate for the effects of the unknown external time-varying disturbance. Due to the fact that z d  jz jD shown in (26), such compensation is achieved by n n handling its bound D. Furthermore, to employ sg(z ; ) in the control input for the compensation, we choose function at (t) = e > 0 where a > 0. Since this function satisfies ( )d < 1; 8t  0, it, together with other parts of the proposed adaptive controller including the estimated D obtained from the adaptive law in (31), ensures that perfect asymptotic output tracking is achieved in our work, which is proved in the main result. Also since sg(z ; ) is differentiable, the obtained control signal in (30) is smooth. Remark 3: Different from some existing control schemes such as that in [45], we do not estimate the actual disturbance for achieving the control objectives pointed out in the problem formulation in Section 2. Instead, we estimate the upper bound D of the disturbance. In addition, the adaptive laws in (31), including the law for updating D, and control law (30) are derived from (29) in such a way that (32) is obtained. Our main results in the next subsection are proved based on (32). As seen from the proof, D is not required to converge to the true unknown bound D and, in fact, such a convergence is not one of the control objectives. 3.2 Main Result For the adaptive controller design above, we have the following result. Theorem 1: Consider the closed-loop system consisting of fractional system (10) and adaptive controller with control law (30) and adaptive laws (15), (21) and (31). The system is globally stable in the sense that all signals in the closed-loop system are uniformly bounded and also, asymptotic tracking is achieved, i.e. lim [y(t) r(t)] = 0. t!1 Proof: Taking fractional integration of both sides of inequality (32) gives C 2 C C V (t)  I ( c z ) + I (D) + V (0)  I (D) + V (0): n j j n n (33) t t t t t t 0 0 0 j=1 In (33), the first term on the right hand side can be computed as follow: Z Z t t a D ( ) D e (34) I (D) = d = d = Dt E (at): 1; +1 t t 1 1 ( ) (t  ) ( ) (t  ) 0 0 According to Lemma 2, sincejarg(at)j =  andjatj ! 1 for t ! 1, by choosing integer n = 1, then we have 1 1 (at) a 1 E (at) = + o(jatj ) = + o : (35) 1; +1 ( + 1 1) ( )t jatj Hence following (35), a 1 t E (at) = + t o : (36) 1; +1 ( )t jatj As t ! 1, the right hand side of (36) tends to 0. Therefore we have lim t E (at) = 0, which t!1 1; +1 implies lim I (D) is also 0. Hence from Lemma 4, we get lim (D) = 0. Finally we can come to the t!1 t!1 t t conclusion that V (t) is bounded from (33), thus every signal in V (t) is bounded. Then all the virtual control signals n n (j = 1; 2; : : : ; n 1) are bounded in accordance with (12) and (20) and further x (i = 1; 2; : : : ; n) are also bounded. j i 7 Besides, from (30) it can be noticed that u is bounded. Therefore, from system (10) it can be shown that the derivatives of x exist and are bounded, which reveals that x are uniformly continuous. i i Since r is uniformly continuous according to Assumption 1, hence z = x r is also uniformly continuous. As a 1 1 result, is uniformly continuous from (12) and (15). Subsequently, as z = x (i = 2; : : : ; n), z can be proved 1 i i i1 i C 2 recursively that are uniformly continuous. From (33) we know that I ( c z ) is bounded. According to Lemma j j t t 0 j=1 5, we have lim z = 0; j = 1; 2; ; n. Therefore the output y(t) tracks reference signal r(t) asymptotically. t!1 j Remark 4: If the control objective is to globally asymptotically stabilize the system, it could be ensured with the same procedure by treating r(t) = 0. In addition, if = 1, then the corresponding results in this paper become those for integer-order case. 4 Simulation Results In this section a second-order and a third-order fractional nonlinear systems are presented as examples for demon- strating and comparing the proposed control method with an existing scheme in [21]. 4.1 A Second-order Example The system to be controlled is given as follows > D x (t) = x (t) +  (x (t)) 1 2 1 1 1 < 0 t (37) D x (t) = u(t) +  (x (t); x (t)) + d(t) 2 2 1 2 2 0 t y(t) = x (t) x 0:5x 2 1 where = 0:95;  (x ) = 0:4x ;  (x ; x ) = 0:1x + . Here  =  = 1 and external time-varying 1 1 2 1 2 2 4 1 2 1+x disturbance d(t) = sin(t) + cos(t) + 2U(t 15) where U(t) is the unit step function are used for simulation purpose, but unknown to designer. The term 2U(t 15) implies that disturbance d(t) has a step jump at t = 15s and thus it does not have bounded derivative. Firstly, the system is simulated when no control is applied, i.e. u(t) = 0, and the result is given in Fig. 1. As observed from the figure, in this case, the system (37) is unstable in the presence of disturbance d(t). To stabilize the system, we apply our proposed adaptive control scheme presented in Section 3. The designed control parameters are selected as c = 30; c = 1; = = 2I;  = 4;  = 2. The responses in this case are shown in Fig. 2 to Fig. 4. It could be 1 2 1 2 1 found from Fig. 2 that the system output y and the virtual error z are driven to 0 eventually. Fig. 3 and Fig. 4 show the ^ ^ ^ estimations of uncertain parameters  and  and bound of external disturbance D as well as the bound of  (t) v;1 v;2 1 accordingly. The control signal u is shown in Fig. 5. To better illustrate the effectiveness of our proposed control algorithm, a comparative simulation study between the schemes in [21] and this paper is conducted under the same control objective that the output signal tracks a reference signal r(t) = sin(0:5t) while ensuring all the signals bounded. In [21], fuzzy logic systems are employed to approximate unknown compounded nonlinear functions in the systems and also the fractional-order derivatives of the virtual control functions. The sign function is used to compensate for the effects caused by system uncertainties and approximated errors, thus causing chattering phenomenon. The comparison results are given in Fig. 6 to Fig. 8, from which we can observe that our proposed method can guarantee both the tracking error z and virtual error z 1 2 converge to 0 asymptotically without leading to chattering phenomenon in control signal u. On the contrary, not only the tracking error z cannot go to 0, but also chattering phenomena in the control signal u and z are caused by using 1 2 control signal in [21], which is also consistent with its theoretical results established. Furthermore, it is proposed in [21] that the chattering phenomenon can be avoided by replacing sign() with arctan(10) in the controller design, without theoretical analysis on stability and tracking performance provided. Comparison simulations are also carried out with such a replacement and the results can be found in Fig. 9 to Fig. 11. It can be seen from these figures that although the control signal in [21] becomes smooth, the tracking error z and the virtual error z can only be driven to small regions 1 2 near 0. However, by utilizing the control method proposed in this paper, both z and z can be asymptotically stabilized 1 2 under smooth control signal u with similar magnitude. 8 20 4 z2 0 3.5 -20 3 -40 2.5 2 -60 2 -80 1.5 -100 1 -1 0 0.5 1 1.5 2 2.5 3 -120 0.5 -140 0 -160 -0.5 0 2 4 6 8 10 12 0 10 20 30 40 50 60 70 80 90 100 Time (seconds) Time (seconds) Figure 1: The state variables of uncontrolled system (37). Figure 2: The system output y = x = z and virtual 1 1 error z . 1.6 8 0.9 ˆ ˆ θ D v,1 ˆ ρˆ θ (2) 1 1.5 v,2 0.85 θv,2(1) 1.4 0.8 1.3 0.75 1.2 0.7 0.65 1.1 0.6 3 0.55 0.9 2 0 0.5 1 1.5 2 2.5 3 0.8 0.7 0.6 -1 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Time (seconds) Time (seconds) ^ ^ ^ Figure 3: The estimates  and  . Figure 4: The estimates D and  ^ . v;1 v;2 1 4.2 An Example of Third-order System Consider the following third-order fractional system which is also known as Chua-Hartley’s system [48]. D x (t) = x (t) +  (x (t)) > 1 2 1 1 1 0 t D x (t) = x (t) +  (x (t); x (t)) 2 3 2 1 2 2 0 t (38) D x (t) = u(t) +  (x (t); x (t); x (t)) + d(t) 3 3 1 2 3 3 0 t y(t) = x (t) 10 100 where = 0:98;  (x ) = (x x );  (x ; x ) = 10x x ;  (x ; x ; x ) = x . Suppose that 1 1 1 2 1 2 1 2 3 1 2 3 2 7 7 =  =  = 1 and external time-varying disturbance d(t) = 0:5sin(2t) + 3U(t 10), but they are unknown to 1 2 3 controller design. The designed control parameters are selected as c = c = c = 2; = = = 0:1I;  = 1 2 3 1 2 3 1 = 1;  = 0:1. To make a comparison, we simulate the system without control and with our proposed control, under the same initial condition x(0) = [0:8; 2; 1]. The behaviour of its state variables are shown in Fig. 12 and Fig. 13, where the green dot indicates the initial state, respectively. As observed from Fig. 12, the system without control exhibits chaotic phenomenon. From Fig. 13, it is seen that our proposed control enables the chaotic behaviors of the original uncontrolled system to be removed. Moreover, from Fig. 13 and Fig. 14, the virtual errors as well as all the states will finally be driven to 0 with control signal u shown in Fig. 15, which also reveals that the system is asymptotically stabilized. All these results illustrate the effectiveness of the proposed schemes and verify our theoretical results established. 9 5 10 u u with control scheme in [21] u with our proposed shceme 0 0 -5 -5 -10 -15 -10 -10 -20 -20 -25 -30 -15 -30 -40 0 0.5 1 1.5 2 2.5 3 -35 -20 -40 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Time (seconds) Time (seconds) Figure 5: The control input u. Figure 6: The control input u with control scheme involv- ing sign() in [21] and with scheme in this paper. 0.15 12 z1 with control scheme in [21] z2 with control scheme in [21] z1 with our proposed shceme z2 with our proposed shceme 0.1 -5 0.05 -10 0 0.5 1 1.5 2 2.5 3 -2 -0.05 -4 -0.1 -6 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Time (seconds) Time (seconds) Figure 7: The output tracking error z with control Figure 8: The virtual error z with control scheme involv- 1 2 scheme involving sign() in [21] and with scheme in this ing sign() in [21] and with scheme in this paper. paper. 10 0.15 u with control scheme in [21] z1 with control scheme in [21] u with our proposed shceme z1 with our proposed shceme 0 0.1 -5 -10 0.05 -15 -20 0 -25 -30 -0.05 -35 -40 -0.1 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Time (seconds) Time (seconds) Figure 9: The control input u with control scheme involv- Figure 10: The output tracking error z with control ing arctan(10) in [21] and with scheme in this paper. scheme involving arctan(10) in [21] and with scheme in this paper. 10 12 40 z2 with control scheme in [21] z2 with our proposed shceme 10 20 8 0 6 -20 4 -40 -2 0 -4 -2 1.5 x 0.5 -6 2 -4 -0.5 0 10 20 30 40 50 60 70 80 90 100 -1 -1.5 Time (seconds) x Figure 11: The virtual error z with control scheme in- Figure 12: The state variables of uncontrolled system (38) volving arctan(10) in [21] and with scheme in this paper. with green dot and red dot representing initial condition and the origin respectively. 0.6 0 0.4 0.2 -10 -20 -0.2 2 -0.4 9 10 11 12 13 14 15 -2 -1 -4 -2 -6 0.8 0.7 0.6 0.5 0.4 2 0.3 -8 0.2 0.1 -3 0 10 20 30 40 50 60 70 80 90 100 -0.1 -0.2 Time (seconds) Figure 13: The state variables of (38) under proposed Figure 14: The virtual errors z and z . 2 3 control scheme with green dot and red dot representing initial condition and the origin respectively. -2 -4 -6 9 10 11 12 13 14 15 -20 -40 -60 0 10 20 30 40 50 60 70 80 90 100 Time (seconds) Figure 15: The control input u of (38). 5 Conclusion In this paper, we propose a smooth adaptive backstepping control design scheme for a class of SISO commensurate fractional-order nonlinear systems in strict feedback form with uncertain system parameters and unknown external 3 time-varying disturbance. It is proved that the resulting closed-loop system is globally asymptotically stable, even in the presence of arbitrary uncertainties and bounded disturbances. Simulation results also demonstrate the effectiveness in stabilizing unstable system and tracking reference signal with better performances compared to an existing scheme presented in [21]. For fractional order > 1, the control protocol in this paper is not theoretically shown effective, and therefore it is an interesting future research topic to extend our result to such systems. References [1] A Anatolii Aleksandrovich Kilbas, Hari Mohan Srivastava, and Juan J Trujillo. Theory and Applications of Fractional Differential Equations, volume 204. Amsterdam, Boston: Elsevier, 2006. [2] Igor Podlubny. Fractional-order systems and fractional-order controllers. Institute of Experimental Physics, Slovak Academy of Sciences, Kosice, 12(3):1–18, Nov. 1994. [3] Peter J Torvik and Ronald L Bagley. On the appearance of the fractional derivative in the behavior of real materials. Journal of Applied Mechanics, 51(2):294–298, Jun. 1984. [4] Michele Caputo and Francesco Mainardi. A new dissipation model based on memory mechanism. Pure and Applied Geophysics, 91(1):134–147, Dec. 1971. [5] CHR Friedrich. Relaxation and retardation functions of the Maxwell model with fractional derivatives. Rheologica Acta, 30(2):151–158, Mar. 1991. [6] Raoul R Nigmatullin and Stuart O Nelson. Recognition of the “fractional” kinetics in complex systems: dielectric properties of fresh fruits and vegetables from 0.01 to 1.8 GHz. Signal Processing, 86(10):2744–2759, Oct. 2006. [7] RC Koeller. Applications of fractional calculus to the theory of viscoelasticity. Journal of Applied Mechanics, 51(2):299–307, Jun. 1984. [8] Juan Yu, Cheng Hu, and Haijun Jiang. -stability and -synchronization for fractional-order neural networks. Neural Networks, 35:82–87, Nov. 2012. [9] JC Wang. Realizations of generalized Warburg impedance with RC ladder networks and transmission lines. Journal of the Electrochemical Society, 134(8):1915–1920, Aug. 1987. [10] Huiyang Liu, Long Cheng, Min Tan, and Zeng-Guang Hou. Exponential finite-time consensus of fractional-order multiagent systems. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 50(4):1549–1558, Apr. 2020. [11] Tom T Hartley and Carl F Lorenzo. Dynamics and control of initialized fractional-order systems. Nonlinear Dynamics, 29(1-4):201–233, Jul. 2002. [12] YangQuan Chen, Hyo-Sung Ahn, and Dingyü Xue. Robust controllability of interval fractional order linear time invariant systems. Signal Processing, 86(10):2794–2802, Jun. 2006. [13] YangQuan Chen, Ivo Petras, and Dingyu Xue. Fractional order control - A tutorial. In 2009 American Control Conference, pages 1397–1411, 2009. [14] Wei Wang, Changyun Wen, and Jing Zhou. Adaptive Backstepping Control of Uncertain Systems with Actuator Failures, Subsystem Interactions, and Nonsmooth Nonlinearities. Boca Raton: CRC Press, 2017. [15] Dumitru Baleanu, José António Tenreiro Machado, and Albert CJ Luo. Fractional Dynamics and Control. New York: Springer Science & Business Media, 2011. [16] Yiheng Wei, Yuquan Chen, Shu Liang, and Yong Wang. A novel algorithm on adaptive backstepping control of fractional order systems. Neurocomputing, 165:395–402, Oct. 2015. [17] Dian Sheng, Yiheng Wei, Songsong Cheng, and Yong Wang. Adaptive backstepping state feedback control for fractional order systems with input saturation. IFAC-PapersOnLine, 50(1):6996–7001, Jul. 2017. [18] Yiheng Wei, W Tse Peter, Zhao Yao, and Yong Wang. Adaptive backstepping output feedback control for a class of nonlinear fractional order systems. Nonlinear Dynamics, 86(2):1047–1056, Jul. 2016. [19] Shabnam Pashaei and Mohammadali Badamchizadeh. A new fractional-order sliding mode controller via a nonlinear disturbance observer for a class of dynamical systems with mismatched disturbances. ISA Transactions, 63:39–48, Jul. 2016. [20] Mou Chen, Shu-Yi Shao, Peng Shi, and Yan Shi. Disturbance-observer-based robust synchronization control for a class of fractional-order chaotic systems. IEEE Transactions on Circuits and Systems II: Express Briefs, 64(4):417–421, Apr. 2017. [21] Heng Liu, Yongping Pan, Shenggang Li, and Ye Chen. Adaptive fuzzy backstepping control of fractional-order nonlinear systems. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 47(8):2209–2217, Aug. 2017. 12 [22] Shuai Song, Baoyong Zhang, Jianwei Xia, and Zhengqiang Zhang. Adaptive backstepping hybrid fuzzy sliding mode control for uncertain fractional-order nonlinear systems based on finite-time scheme. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 50(4):1559–1569, Apr. 2020. [23] Zhen Wang. Synchronization of an uncertain fractional-order chaotic system via backstepping sliding mode control. Discrete Dynamics in Nature and Society, 2013, Jun. 2013. [24] Ping Gong and Weiyao Lan. Adaptive robust tracking control for multiple unknown fractional-order nonlinear systems. IEEE Transactions on Cybernetics, 49(4):1365–1376, Apr. 2019. [25] Qiao Wang, Jianliang Zhang, Dongsheng Ding, and Donglian Qi. Adaptive Mittag-Leffler stabilization of a class of fractional order uncertain nonlinear systems. Asian Journal of Control, 18(6):2343–2351, Nov. 2016. [26] Dongsheng Ding, Donglian Qi, and Qiao Wang. Non-linear Mittag–Leffler stabilisation of commensurate fractional-order non-linear systems. IET Control Theory & Applications, 9(5):681–690, Mar. 2015. [27] Dongsheng Ding, Donglian Qi, Yao Meng, and Li Xu. Adaptive Mittag-Leffler stabilization of commensurate fractional-order nonlinear systems. In 53rd IEEE Conference on Decision and Control, pages 6920–6926, 2014. [28] Zhiyao Ma and HongJun Ma. Adaptive fuzzy backstepping dynamic surface control of strict-feedback fractional order uncertain nonlinear systems. IEEE Transactions on Fuzzy Systems, 28(1):122–133, Jan. 2020. [29] Shuai Song, Baoyong Zhang, Xiaona Song, and Zhengqiang Zhang. Neuro-fuzzy-based adaptive dynamic surface control for fractional-order nonlinear strict-feedback systems with input constraint. IEEE Transactions on Systems, Man, and Cybernetics: Systems, to be published, doi: 10.1109/TSMC.2019.2933359. [30] Igor Podlubny. Fractional Differential Equations: An Introduction to Fractional Derivatives, Fractional Differen- tial Equations, to Methods of their Solution and some of their Applications, volume 198. San Diego: Academic Press, 1999. [31] Dingyü Xue. Fractional-Order Control Systems: Fundamentals and Numerical Implementations, volume 1. Berlin: De Gruyter, 2017. [32] V. Lakshmikantham, Srinivasa Leela, and J. Vasundhara Devi. Theory of Fractional Dynamic Systems. Cambridge: CSP, 2009. [33] Changpin Li and Weihua Deng. Remarks on fractional derivatives. Applied Mathematics and Computation, 187(2):777–784, Apr. 2007. [34] Bijnan Bandyopadhyay and Shyam Kamal. Stabilization and Control of Fractional Order Systems: A Sliding Mode Approach, volume 317. Cham: Springer, 2015. [35] Norelys Aguila-Camacho, Manuel A Duarte-Mermoud, and Javier A Gallegos. Lyapunov functions for fractional order systems. Communications in Nonlinear Science and Numerical Simulation, 19(9):2951–2957, Sep. 2014. [36] Yan Li, YangQuan Chen, and Igor Podlubny. Mittag–Leffler stability of fractional order nonlinear dynamic systems. Automatica, 45(8):1965–1969, Aug. 2009. [37] Yan Li, YangQuan Chen, and Igor Podlubny. Stability of fractional-order nonlinear dynamic systems: Lyapunov direct method and generalized Mittag–Leffler stability. Computers & Mathematics with Applications, 59(5):1810– 1821, Mar. 2010. [38] Fengrong Zhang, Changpin Li, and YangQuan Chen. Asymptotical stability of nonlinear fractional differential system with Caputo derivative. International Journal of Differential Equations, 2011, Aug. 2011. [39] Javier A Gallegos, Manuel A Duarte-Mermoud, Norelys Aguila-Camacho, and Rafael Castro-Linares. On fractional extensions of Barbalat lemma. Systems & Control Letters, 84:7–12, Oct. 2015. [40] Ruoxun Zhang and Yongli Liu. A new Barbalat’s lemma and Lyapunov stability theorem for fractional order systems. In 2017 29th Chinese control and decision conference (CCDC), pages 3676–3681, 2017. [41] Zongyu Zuo and Chenliang Wang. Adaptive trajectory tracking control of output constrained multi-rotors systems. IET Control Theory & Applications, 8(13):1163–1174, Sep. 2014. [42] Dong Yang, Guangdeng Zong, and Hamid Reza Karimi. H refined anti-disturbance control of switched LPV systems with application to aero-engine. IEEE Transactions on Industrial Electronics, 67(4):3180–3190, Apr. [43] Guangdeng Zong, Yankai Li, and Haibin Sun. Composite anti-disturbance resilient control for Markovian jump nonlinear systems with general uncertain transition rate. Science China Information Sciences, 62(2):22205, Jan. [44] Haibin Sun, Yankai Li, Guangdeng Zong, and Linlin Hou. Disturbance attenuation and rejection for stochastic Markovian jump system with partially known transition probabilities. Automatica, 89:349–357, Mar. 2018. 13 [45] Huaguang Zhang, Jian Han, Chaomin Luo, and Yingchun Wang. Fault-tolerant control of a nonlinear system based on generalized fuzzy hyperbolic model and adaptive disturbance observer. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 47(8):2289–2300, Aug. 2017. [46] Guy Jumarie. Modified Riemann-Liouville derivative and fractional Taylor series of nondifferentiable functions further results. Computers & Mathematics with Applications, 51(9-10):1367–1376, May 2006. [47] Jian Wang, Yanqing Wen, Yida Gou, Zhenyun Ye, and Hua Chen. Fractional-order gradient descent learning of BP neural networks with Caputo derivative. Neural Networks, 89:19–30, May 2017. [48] Ivo Petráš. A note on the fractional-order Chua’s system. Chaos, Solitons & Fractals, 38(1):140–147, Oct. 2008. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Electrical Engineering and Systems Science arXiv (Cornell University)

Adaptive Backstepping Control for Fractional-Order Nonlinear Systems with External Disturbance and Uncertain Parameters Using Smooth Control

Download PDF
14 pages

Loading next page...
 
/lp/arxiv-cornell-university/adaptive-backstepping-control-for-fractional-order-nonlinear-systems-Vs7xI9K8X5

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

ISSN
2168-2216
eISSN
ARCH-3348
DOI
10.1109/TSMC.2020.2987335
Publisher site
See Article on Publisher Site

Abstract

Adaptive Backstepping Control for Fractional-Order Nonlinear Systems with External Disturbance and Uncertain Parameters Using Smooth Control Xinyao Li Changyun Wen School of Electrical and Electronic Engineering School of Electrical and Electronic Engineering Nanyang Technological University Nanyang Technological University Singapore, 639798 Singapore, 639798 E180209@e.ntu.edu.sg ecywen@ntu.edu.sg Ying Zou School of Electrical and Electronic Engineering Nanyang Technological University Singapore, 639798 zouy0011@e.ntu.edu.sg April 15, 2020 ABSTRACT In this paper, we consider controlling a class of single-input-single-output (SISO) commensurate fractional-order nonlinear systems with parametric uncertainty and external disturbance. Based on backstepping approach, an adaptive controller is proposed with adaptive laws that are used to estimate the unknown system parameters and the bound of unknown disturbance. Instead of using discontinuous functions such as the sign function, an auxiliary function is employed to obtain a smooth control input that is still able to achieve perfect tracking in the presence of bounded disturbances. Indeed, global boundedness of all closed-loop signals and asymptotic perfect tracking of fractional-order system output to a given reference trajectory are proved by using fractional directed Lyapunov method. To verify the effectiveness of the proposed control method, simulation examples are presented. Keywords Adaptive backstepping control; fractional-order; nonlinear systems; smooth control. 1 Introduction Fractional calculus owns a history of more than 300 years. It is a branch of mathematics that deals with non-integer order derivatives and integrals. Compared with integer order calculus, fractional order integral and derivative can both be treated as weighted integral and thus they have the properties of hereditary and infinite memory [1, 2], which can also be seen from their definitions given in Section 2. Such properties give the most significant meaning of fractional-order derivative compared to integer-order derivative that does not have such properties. Thus, using fractional order models could better and more accurately describe the characteristics of the real world systems than integer order models, as elaborated in [3–5]. In the past few years, many researchers have paid significant attention to fractional order calculus and constructed models for real-world systems, including viscoelasticity, complex systems, neural networks, transmission line, multi agent systems, and so on, see for examples [6–10]. Moreover, fractional order systems as well as their controls have also been studied in recent years [11–13]. However, it is difficult to simply apply the approaches of controller design and analysis developed for integer-order systems to fractional-order systems due to the lack of appropriate mathematical tools. For example, the fractional-order derivative of a composite function is the sum of infinite number of terms, which is different from the concise closed form expression of its integer-order derivative easily obtained by applying the chain rule. Backstepping technique, which demonstrates a step-by-step design procedure by constructing Lyapunov function and virtual control signal at each step, is widely applied in controlling integer-order nonlinear systems, see [14] and [15] for arXiv:2002.00162v3 [eess.SY] 14 Apr 2020 example. However, it is not easy to directly employ backstepping control technique on fractional-order systems due to the challenge of obtaining the fractional derivative of the quadratic-type Lyapunov function. In [15], the backstepping technique is extended to fractional order systems without entirely considering uncertainties. Integer-order Lyapunov ( ) method is applied in [15] by proving zz _ < 0 if zz < 0, where z denotes virtual error defined in backstepping algorithm and 0 < < 1. A method that transforms backstepping control problem for the fractional-order systems to integer-order by taking into account the frequency distributed model is proposed in [16–18]. In this way, the stability analysis for fractional-order systems is also carried out with integer-order Lyapunov method. In order to handle disturbances in fractional-order nonlinear systems, nonlinear disturbances observers are designed to counteract the effects caused by unknown disturbances in [19] and [20]. However, strong assumptions that the disturbances must be constant or they should have bounded derivatives should be met to achieve satisfactory perfor- mances. Adaptive controllers utilizing sign function are developed in [21–24], which ensure the stabilization/tracking error converges to zero with non-continuous control because of the non-continuity of sign function. For the purpose of avoiding calculating the fractional-order derivatives of virtual control signals, the involved derivatives are directly subtracted in the design of virtual controllers in subsequent steps in [25–27]. However, the boundedness of signals in the closed-loop control systems is not theoretically shown in these works. Fuzzy logic is employed in [21] and [28] to deal with system uncertainties as well as computing fractional order derivatives of virtual control signals with the assumption that the errors resulted from fuzzy approximation of the true values of system uncertainties and the fractional order derivatives of virtual control signals are bounded. In addition, both works can only show that the output tracking errors tend to an arbitrary small region. Besides, [28] combines dynamic surface control with adaptive backstepping method for SISO fractional-order nonlinear systems with uncertain system functions and multiple external unknown disturbances. Although it can eliminate the chattering phenomenon in control signals by getting rid of using the sign function, their results can only be obtained if the initial conditions are located in certain range. Dynamic surface control integrated with neuro-fuzzy network for fractional-order nonlinear system subjects to input constraint is addressed in [29]. Similarly, the output tracking error is only guaranteed to converge to a small region around the origin with constraint of the initial conditions. Therefore, it is still open and challenging to construct an adaptive backstepping controller with smooth control signal for fractional nonlinear systems involving both parametric uncertainties and external time-varying disturbance, which ensures asymptotic stability and perfect tracking property without restrictions on the initial conditions. Inspired by the discussions above, in this paper we address such an issue. To have a smooth control signal, an auxiliary function is used in lieu of the sign function in the designed controller. We employ the fractional directed Lyapunov method for our design and analysis. As we know, when backstepping approach is employed to design controller for high-order nonlinear systems, the time derivatives of virtual control signals are required and they can easily be obtained for integer-order systems by chain rule. But, as mentioned above, this is not the case for fractional-order systems. To overcome this difficulty, a novel approach for approximating the fractional-order time derivatives of virtual control signals is proposed and adaptive laws are designed to estimate the bounds of approximation errors. It is shown that, with the designed controller and adaptive laws, the closed loop system is globally stable and its output tracks a given reference input asymptotically, even in the presence of external time-varying disturbance and uncertain parameters. To the best of the authors’ knowledge, this is the first paper to have such results. Simulation studies illustrate the effectiveness of the proposed control scheme and also reveal its advantages compared to an existing approach. In summary, the main contribution of this paper is to design a smooth control that achieves asymptotic perfect output tracking/stabilization for fractional-order nonlinear commensurate systems and ensures global stability in the sense that all signals in the closed-loop systems remain globally bounded, even in the presence of unknown bounded external disturbances and also system uncertainties. The rest of the paper is organized as follows. Preliminaries and the fractional-order system description are provided in Section 2. In Section 3, the design of an adaptive controller is presented in detail. In Section 4, the scheme is illustrated by simulation studies with comparison to that in [21]. Finally the paper is concluded in Section 5. 2 Preliminaries and Problem Formulation 2.1 Preliminaries Definition 1 [30]: The fractional integral of an integrable function f (t) with 2 R is 1 f ( ) I f (t) = d (1) 0 t ( ) (t  ) where I means the fractional integral of order with initial time t and () denotes the well-known Gamma t 0 0 t t z1 function, which is defined as (z) = e t dt, where z 2 C. One of the significant properties of Gamma 2 function is [31]: (z + 1) = z(z); (n + 1) = n(n) = n(n 1)(n 1) =  = n!; (n) = 1, where n 2 N = fnjn  0; n 2 Ng. Definition 2 [30]: The Caputo fractional derivative of a function is shown as m (m) d 1 f ( ) (m ) D f (t) = I f (t) = d (2) t t 0 t 0 m m+1 dt (m ) (t  ) where m 1 < < m 2 Z . From equation (2) we can observe that the Caputo derivative of a constant is 0. Another commonly used fractional derivative is named Riemann-Liouville (RL) and the RL fractional derivative of a function RL f (t) is denoted as D f (t). Different from the Caputo derivative, RL derivative of a constant is not equal to 0 [30, 32]. t t To obtain the unique solution for fractional differential equation D x(t) = f (x; t), (m 1 < < m 2 0 t Z and t  t ), the initial values need to be determined. According to [30, 33] and [34], fractional differential equations with Caputo-type derivative have initial values that are in-line with integer-order differential equations, i.e. 0 (m1) x(t ); x (t ); : : : ; x (t ), which contain specific physical interpretations. On the contrary, although mathematically 0 0 0 the initial value problem for RL fractional differential equations is rigorous and solvable, it lacks of practical explanation since the physical meanings of these initial conditions are unknown yet. Therefore, Caputo-type fractional systems are frequently employed in practical analysis. n + Lemma 1 [35]: Assume x(t) 2 R is a smooth function, n 2 Z , then for 8t  t , C T T C D [x (t)Px(t)]  x (t)P D x(t); 0 < < 1 (3) t t t t 0 0 nn where P 2 R is a positive definite constant matrix. Lemma 2 [30]: If 0 < < 2, represents an arbitrary complex number and real number  satisfies that <  < minf;  g, then for any integer n  1, 1n E (z) = + o(jzj ); (4) ( j) j=1 where   jarg(z)j   and jzj ! 1. Here E (z) is the Mittag-Leffler function, which is described as follows: E (z) = ; ( > 0; > 0): (5) ( k + ) k=0 Lemma 3 [33, 35]: If the Caputo fractional derivative D f (t) is integrable, then t t m1 (k) f (t ) C k I D f (t) = f (t) (t t ) (6) t 0 0 t t t k! k=0 where m 1 < < m 2 Z . Particularly, for 0 <  1, we can obtain I D f (t) = f (t) f (t ): (7) t 0 0 t t t Definition 3 [36–38]: For fractional nonautonomous system D x (t) = f (x; t), where i = 1; 2; : : : ; n, 0 < < 1, t i i 0 t T n initial condition is x(t ) = [x (t ); x (t ); : : : ; x (t )] 2 R , D indicates Caputo or RL fractional derivative, 0 1 0 2 0 n 0 t 0 t f (x; t) : [t ;1) ! R is locally Lipschitz in x and piecewise continuous in t (which insinuates the existence i 0 and uniqueness of the solution to the fractional systems [30]) and 2 R stands for a region that contains the origin T     T x = [0; 0; : : : ; 0] . The equilibrium x = [x ; x ; : : : ; x ] of this system is defined as D x = f (x ; t) for t  t . t i 0 1 2 n 0 t Without loss of generality, we set the equilibrium x = [0; 0; : : : ; 0] . + + Lemma 4 [39]: Assume f (t) : R ! R is uniformly continuous and lim I f (t) = 0 with 0 < < 1 and t!1 t 0 t 8t > t > 0, then lim f (t) = 0. 0 t!1 n n Lemma 5 [40]: If x(t) : [t ;1)  R ! R is uniformly continuous, I jx(t)j  K for 8t > t > 0 and 0 t 0 0 t 0 < < 1 with constants p; K > 0, then lim x(t) = 0. t!1 Lemma 6 [41]: The following inequality holds for x 2 R: jxj <  + xsg(x; ) (8) 3 where sg(x; ) = p (9) 2 2 x + and  is a differentiable function which meets  > 0 and ( )d < 1; 8t  0. From equation (9), it can be derived that sg(x; ) is differentiable and thus it is a smooth function for8t  0. 2.2 Problem Formulation In this paper, Caputo-type definition of the fractional derivatives is utilized. Consider the following class of n-th order SISO uncertain commensurate fractional-order nonlinear systems with time-varying disturbance in strict feedback form: C T D x (t) = x (t) +  (x (t)) > 1 2 1 1 0 t 1 C T D x (t) = x (t) +  (x (t)) > 2 3 2 2 0 t 2 < . (10) C T D x (t) = x (t) +  (x (t)) n1 n n1 n1 0 t n1 C T D x (t) = g(x(t))u(t) +  (x(t)) + d(t) n n 0 t n y(t) = x (t) where the fractional-orders of all the state equations are equal to 2 (0; 1], u(t) 2 R and y(t) 2 R represent the control T n input and system output respectively, x(t) = [x (t); x (t); ; x (t)] 2 R denotes the measurable state vector, 1 2 n T i p x (t) = [x (t); x (t); ; x (t)] 2 R ,  (x (t)) 2 R (i = 1; 2; ; n) is a vector with its elements being known i 1 2 i i i nonlinear smooth functions,  2 R (i = 1; 2; ; n) is an unknown constant vector, g(x(t)) is a known non-zero smooth nonlinear function and d(t) stands for an unknown bounded time-varying external disturbance with unknown bound D. Remark 1: Different from the control strategies in [42–45] that are designed with disturbance observer or disturbance resilience, a fractional order adaptive law, which can be found in (31), is to be derived to estimate the unknown upper bound D of the disturbance. The external time-varying disturbance d(t) is allowed to be norm-bounded without any condition on its derivative in this paper. By employing the estimated bound in designing control input u, such unknown disturbance can be compensated, which helps us to solve the following control problem. The control problem is to design an adaptive controller for the class of systems described in (10) such that the following objectives are achieved: 1) the closed-loop system is globally stable in the sense that all the signals including parameter estimates are bounded; 2) the system output y(t) asymptotically tracks a reference signal r(t), i.e. lim [y(t) r(t)] = 0. t!1 Assumption 1: The given reference signal r(t) and its -th order Caputo-type fractional derivative D r(t) are 0 t smooth and bounded. 3 Adaptive Controller Design and Stability Analysis 3.1 Adaptive Controller Design To achieve the above objectives, an adaptive controller is designed based on the backstepping design procedure. Defining virtual errors as follows: z = x r, which is also known as tracking error, z = x (i = 2; : : : ; n), 1 1 i i i1 where (j = 1; 2; : : : ; n 1) denotes the virtual control signal that will be designed iteratively later. 1 2 Step 1: Consider the first Lyapunov function V = z . According to Lemma 1, the Caputo derivative of V is 1 1 1 C C 2 C C C T C D V = D z  z D z = z [ D x D r] = z [x +  (x ) D r] 1 1 1 1 1 1 1 2 1 1 0 t 0 t 0 t 0 t 0 t 1 0 t (11) T C T C = z [z + +  (x ) D r] = z z + z + z  (x ) z D r: 1 2 1 1 1 1 2 1 1 1 1 1 1 1 0 t 1 0 t Let virtual control signal be T C = c z  (x ) + D r; (12) 1 1 1 1 1 1 0 t 4 ^ ~ ^ where  stands for the estimates of  and c is a positive design parameter. Defining  =   , then we have 1 1 1 1 1 1 C C ~ ^ D  = D  . Therefore, inequality (11) becomes 1 1 0 t 0 t C T C 2 T C T C D V z z + z + z  (x ) z D r = z z c z z  (x ) + z D r + z  (x ) z D r 1 1 2 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 0 t 1 0 t 1 0 t 1 0 t 2 T =z z c z + z  (x ) : 1 2 1 1 1 1 1 (13) 1 T 1 ~ ~ Let V = V +   , where is positive definite matrix, then the Caputo derivative of V is 1 1 1 1 1 2 1 C C C T 1 C T 1C 2 T T 1C ~ ~ ~ ~ ~ ~ ~ D V = D V + D    D V +  D   z z c z + z  (x ) +  D 1 1 1 1 1 1 2 1 1 1 1 1 1 0 t 0 t 0 t 1 1 0 t 1 1 0 t 1 1 1 0 t 2 T T C ~ ~ ^ =z z c z + z  (x )  D  : 1 2 1 1 1 1 1 1 1 1 1 0 t (14) Designing adaptive law as D  = z  (x ); (15) 1 1 1 1 1 0 t then inequality (14) becomes C 2 T T 1 2 ~ ~ (16) D V z z c z + z  (x )   (x )z = z z c z : 1 1 2 1 1 1 1 1 1 1 1 1 1 2 1 1 0 t 1 1 Step i (i = 2; ; n 1): The fractional order derivative of z is C C C T C T C D z = D x D = x +  (x (t)) D = z + +  (x (t)) D : (17) i i i1 i+1 i i i1 i+1 i i i i1 0 t 0 t 0 t i 0 t i 0 t Inspired by the idea in [46] and [47] where D f [u(x)] is approximated by f (u) D u(x) with approxima- 0 x u 0 x i1 @ C i1 C tion error (t) which is bounded by an unknown bound  , D is approximated by ( ) D x + i1 j 0 t 0 t j=1 @x i1 @ i1 C ( ) D  . Therefore by defining  (t) as the approximation error with bound   , (17) can be written v;j i1 i1 0 t j=1 ^ v;j as C T D z = z + +   +  +  : (18) i i+1 i v;i i i1 0 t v;i The functions and parameters in (18) are given below @ @ @ i1 i1 i1 T T T T T =[ (x ); ( ) (x ); ; ( ) (x ); ( ) (x )] ; v;i i i1 2 1 i i1 2 1 @x @x @x i1 2 1 T T T T =[ ;  ; ;  ] ; v;i i i1 1 (19) i i1 X X @ @ i1 i1 = ( )x + ( ) D  ; i j v;j 0 t @x j1 @ v;j j=2 j=1 where  is the estimate of  ; j = 1; 2; ; i 1. v;j v;j Choose virtual control signal as = z c z     ^ ; (20) i i1 i i v;i i i1 v;i where c denotes a positive design parameter,  means the estimate of  and  ^ is the estimation of   with i v;i v;i i1 i1 their adaptive law respectively designed as D  = z  ; v;i i i v;i 0 t (21) D  ^ = z  ; i1 i i1 0 t 1 2 1 T 1 2 ~ ~ ~ ^ where is positive definite and  > 0. Let V = V + z +   +  ~ , where  = i i1 i i1 i v;i v;i v;i v;i v;i i i1 2 2 2 i1 and  ~ =    ^ , then i1 i1 i1 C C C T 1C C ~ ~ D V  D V + z D z +  D  +  ~ D  ~ 0 t i 0 t i1 i0 t i v;i 0 t v;i i10 t i1 i1 (22) C C T 1C C ~ ^ = D V + z D z  D   ~ D  ^ : i1 i i v;i i1 i1 0 t 0 t v;i i 0 t 0 t i1 5 P i1 C 2 Since D V  z z c z , thus we get i1 i1 i j j 0 t j=1 C 2 D V  z z c z : (23) i i i+1 j j 0 t j=1 Step n: The fractional order derivative of z is C T D z = g(x)u + d +   +  +  ; (24) n v;n n n1 0 t v;n where @ @ @ n1 n1 n1 T T T T T =[ (x); ( ) (x ); ; ( ) (x ); ( ) (x )] ; v;n n1 2 1 n n1 2 1 @x @x @x n1 2 1 T T T T =[ ;  ; ;  ] ; v;n n n1 1 n n1 X X @ @ n1 n1 = ( )x + ( ) D  ; n j v;j 0 t @x j1 @ v;j j=2 j=1 (25) n1 n1 X X @ @ n1 n1 C C = ( ) D x + ( ) D n1 j v;j 0 t 0 t @x j @ v;j j=1 j=1 D ; n1 0 t j j : n1 n1 1 2 Considering the Lyapunov function V = V + z , then the Caputo derivative of V is n n1 n n C C C 2 C C D V = D V + D z  D V + z D z n n1 n n1 n n 0 t 0 t 0 t 0 t 0 t n1 2 T z z c z + z [g(x)u + d +   +  +   ] n1 n j j n v;n n n1 v;n (26) j=1 n1 2 T z z c z + z [g(x)u +   +  +   ] +jz j D; n1 n j j n v;n n n1 n v;n j=1 where D is an unknown bound of d(t). Considering Lemma 6, since jz j <  + z sg(z ; ), hence jz j D n n n n D + z sg(z ; )D. Therefore, (26) can be written as n n n1 C 2 T D V z z c z + z [g(x)u +   +  +   ] + D + z sg(z ; )D: (27) n n1 n j j n v;n n n1 n n 0 t v;n j=1 1 1 1 1 2 T 1 2 2 ~ ~ ~ Let V = V + z +   +  ~ + D , where is positive definite,  > 0,  > 0, n n1 n v;n n n1 v;n n n1 2 2 2 2 n1 ~ ^ ~ ^ ^ ^ =   ,  ~ =    ^ , D = D D and  ,  ^ and D represent the estimates of  , v;n v;n v;n n1 n1 n1 v;n n1 v;n n1 and D respectively. Then 1 1 C C C T 1C C C ~ ~ ~ ~ D V  D V + z D z +  D  +  ~ D  ~ + D D D n n1 n n v;n n1 n1 0 t 0 t 0 t v;n n 0 t 0 t 0 t n1 (28) 1 1 C C T 1C C C ~ ^ ~ ^ = D V + z D z  D   ~ D  ^ D D D: 0 t n1 n0 t n v;n n 0 t v;n n10 t n1 0 t n1 According to (26) and (27), we then have n1 C 2 T D V z z c z + z [g(x)u +   +  +   ] + D + z sg(z ; )D n n1 n j j n v;n n n1 n n 0 t v;n j=1 (29) 1 1 T 1C C C ~ ^ ~ ^ D   ~ D  ^ D D D: v;n n1 n1 v;n n 0 t 0 t 0 t n1 Finally the real control input u is designed as ^ ^ u = [z c z   sg(z ; )D   ^ ]; (30) n1 n n v;n n n n1 v;n g(x) 6 at where function (t) = e with a being a positive constant and c is positive design parameter. Also we have the following adaptive laws designed D  = z  ; v;n n n v;n 0 t D  ^ = z  ; (31) n1 n n1 0 t D D = z sg(z ; ): n n 0 t Hence by substituting (30) and (31) into (29), we have C 2 D V  c z + D: (32) n j j 0 t j=1 Remark 2: Function sg(z ; ) in the control law (30) is engaged to compensate for the effects of the unknown external time-varying disturbance. Due to the fact that z d  jz jD shown in (26), such compensation is achieved by n n handling its bound D. Furthermore, to employ sg(z ; ) in the control input for the compensation, we choose function at (t) = e > 0 where a > 0. Since this function satisfies ( )d < 1; 8t  0, it, together with other parts of the proposed adaptive controller including the estimated D obtained from the adaptive law in (31), ensures that perfect asymptotic output tracking is achieved in our work, which is proved in the main result. Also since sg(z ; ) is differentiable, the obtained control signal in (30) is smooth. Remark 3: Different from some existing control schemes such as that in [45], we do not estimate the actual disturbance for achieving the control objectives pointed out in the problem formulation in Section 2. Instead, we estimate the upper bound D of the disturbance. In addition, the adaptive laws in (31), including the law for updating D, and control law (30) are derived from (29) in such a way that (32) is obtained. Our main results in the next subsection are proved based on (32). As seen from the proof, D is not required to converge to the true unknown bound D and, in fact, such a convergence is not one of the control objectives. 3.2 Main Result For the adaptive controller design above, we have the following result. Theorem 1: Consider the closed-loop system consisting of fractional system (10) and adaptive controller with control law (30) and adaptive laws (15), (21) and (31). The system is globally stable in the sense that all signals in the closed-loop system are uniformly bounded and also, asymptotic tracking is achieved, i.e. lim [y(t) r(t)] = 0. t!1 Proof: Taking fractional integration of both sides of inequality (32) gives C 2 C C V (t)  I ( c z ) + I (D) + V (0)  I (D) + V (0): n j j n n (33) t t t t t t 0 0 0 j=1 In (33), the first term on the right hand side can be computed as follow: Z Z t t a D ( ) D e (34) I (D) = d = d = Dt E (at): 1; +1 t t 1 1 ( ) (t  ) ( ) (t  ) 0 0 According to Lemma 2, sincejarg(at)j =  andjatj ! 1 for t ! 1, by choosing integer n = 1, then we have 1 1 (at) a 1 E (at) = + o(jatj ) = + o : (35) 1; +1 ( + 1 1) ( )t jatj Hence following (35), a 1 t E (at) = + t o : (36) 1; +1 ( )t jatj As t ! 1, the right hand side of (36) tends to 0. Therefore we have lim t E (at) = 0, which t!1 1; +1 implies lim I (D) is also 0. Hence from Lemma 4, we get lim (D) = 0. Finally we can come to the t!1 t!1 t t conclusion that V (t) is bounded from (33), thus every signal in V (t) is bounded. Then all the virtual control signals n n (j = 1; 2; : : : ; n 1) are bounded in accordance with (12) and (20) and further x (i = 1; 2; : : : ; n) are also bounded. j i 7 Besides, from (30) it can be noticed that u is bounded. Therefore, from system (10) it can be shown that the derivatives of x exist and are bounded, which reveals that x are uniformly continuous. i i Since r is uniformly continuous according to Assumption 1, hence z = x r is also uniformly continuous. As a 1 1 result, is uniformly continuous from (12) and (15). Subsequently, as z = x (i = 2; : : : ; n), z can be proved 1 i i i1 i C 2 recursively that are uniformly continuous. From (33) we know that I ( c z ) is bounded. According to Lemma j j t t 0 j=1 5, we have lim z = 0; j = 1; 2; ; n. Therefore the output y(t) tracks reference signal r(t) asymptotically. t!1 j Remark 4: If the control objective is to globally asymptotically stabilize the system, it could be ensured with the same procedure by treating r(t) = 0. In addition, if = 1, then the corresponding results in this paper become those for integer-order case. 4 Simulation Results In this section a second-order and a third-order fractional nonlinear systems are presented as examples for demon- strating and comparing the proposed control method with an existing scheme in [21]. 4.1 A Second-order Example The system to be controlled is given as follows > D x (t) = x (t) +  (x (t)) 1 2 1 1 1 < 0 t (37) D x (t) = u(t) +  (x (t); x (t)) + d(t) 2 2 1 2 2 0 t y(t) = x (t) x 0:5x 2 1 where = 0:95;  (x ) = 0:4x ;  (x ; x ) = 0:1x + . Here  =  = 1 and external time-varying 1 1 2 1 2 2 4 1 2 1+x disturbance d(t) = sin(t) + cos(t) + 2U(t 15) where U(t) is the unit step function are used for simulation purpose, but unknown to designer. The term 2U(t 15) implies that disturbance d(t) has a step jump at t = 15s and thus it does not have bounded derivative. Firstly, the system is simulated when no control is applied, i.e. u(t) = 0, and the result is given in Fig. 1. As observed from the figure, in this case, the system (37) is unstable in the presence of disturbance d(t). To stabilize the system, we apply our proposed adaptive control scheme presented in Section 3. The designed control parameters are selected as c = 30; c = 1; = = 2I;  = 4;  = 2. The responses in this case are shown in Fig. 2 to Fig. 4. It could be 1 2 1 2 1 found from Fig. 2 that the system output y and the virtual error z are driven to 0 eventually. Fig. 3 and Fig. 4 show the ^ ^ ^ estimations of uncertain parameters  and  and bound of external disturbance D as well as the bound of  (t) v;1 v;2 1 accordingly. The control signal u is shown in Fig. 5. To better illustrate the effectiveness of our proposed control algorithm, a comparative simulation study between the schemes in [21] and this paper is conducted under the same control objective that the output signal tracks a reference signal r(t) = sin(0:5t) while ensuring all the signals bounded. In [21], fuzzy logic systems are employed to approximate unknown compounded nonlinear functions in the systems and also the fractional-order derivatives of the virtual control functions. The sign function is used to compensate for the effects caused by system uncertainties and approximated errors, thus causing chattering phenomenon. The comparison results are given in Fig. 6 to Fig. 8, from which we can observe that our proposed method can guarantee both the tracking error z and virtual error z 1 2 converge to 0 asymptotically without leading to chattering phenomenon in control signal u. On the contrary, not only the tracking error z cannot go to 0, but also chattering phenomena in the control signal u and z are caused by using 1 2 control signal in [21], which is also consistent with its theoretical results established. Furthermore, it is proposed in [21] that the chattering phenomenon can be avoided by replacing sign() with arctan(10) in the controller design, without theoretical analysis on stability and tracking performance provided. Comparison simulations are also carried out with such a replacement and the results can be found in Fig. 9 to Fig. 11. It can be seen from these figures that although the control signal in [21] becomes smooth, the tracking error z and the virtual error z can only be driven to small regions 1 2 near 0. However, by utilizing the control method proposed in this paper, both z and z can be asymptotically stabilized 1 2 under smooth control signal u with similar magnitude. 8 20 4 z2 0 3.5 -20 3 -40 2.5 2 -60 2 -80 1.5 -100 1 -1 0 0.5 1 1.5 2 2.5 3 -120 0.5 -140 0 -160 -0.5 0 2 4 6 8 10 12 0 10 20 30 40 50 60 70 80 90 100 Time (seconds) Time (seconds) Figure 1: The state variables of uncontrolled system (37). Figure 2: The system output y = x = z and virtual 1 1 error z . 1.6 8 0.9 ˆ ˆ θ D v,1 ˆ ρˆ θ (2) 1 1.5 v,2 0.85 θv,2(1) 1.4 0.8 1.3 0.75 1.2 0.7 0.65 1.1 0.6 3 0.55 0.9 2 0 0.5 1 1.5 2 2.5 3 0.8 0.7 0.6 -1 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Time (seconds) Time (seconds) ^ ^ ^ Figure 3: The estimates  and  . Figure 4: The estimates D and  ^ . v;1 v;2 1 4.2 An Example of Third-order System Consider the following third-order fractional system which is also known as Chua-Hartley’s system [48]. D x (t) = x (t) +  (x (t)) > 1 2 1 1 1 0 t D x (t) = x (t) +  (x (t); x (t)) 2 3 2 1 2 2 0 t (38) D x (t) = u(t) +  (x (t); x (t); x (t)) + d(t) 3 3 1 2 3 3 0 t y(t) = x (t) 10 100 where = 0:98;  (x ) = (x x );  (x ; x ) = 10x x ;  (x ; x ; x ) = x . Suppose that 1 1 1 2 1 2 1 2 3 1 2 3 2 7 7 =  =  = 1 and external time-varying disturbance d(t) = 0:5sin(2t) + 3U(t 10), but they are unknown to 1 2 3 controller design. The designed control parameters are selected as c = c = c = 2; = = = 0:1I;  = 1 2 3 1 2 3 1 = 1;  = 0:1. To make a comparison, we simulate the system without control and with our proposed control, under the same initial condition x(0) = [0:8; 2; 1]. The behaviour of its state variables are shown in Fig. 12 and Fig. 13, where the green dot indicates the initial state, respectively. As observed from Fig. 12, the system without control exhibits chaotic phenomenon. From Fig. 13, it is seen that our proposed control enables the chaotic behaviors of the original uncontrolled system to be removed. Moreover, from Fig. 13 and Fig. 14, the virtual errors as well as all the states will finally be driven to 0 with control signal u shown in Fig. 15, which also reveals that the system is asymptotically stabilized. All these results illustrate the effectiveness of the proposed schemes and verify our theoretical results established. 9 5 10 u u with control scheme in [21] u with our proposed shceme 0 0 -5 -5 -10 -15 -10 -10 -20 -20 -25 -30 -15 -30 -40 0 0.5 1 1.5 2 2.5 3 -35 -20 -40 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Time (seconds) Time (seconds) Figure 5: The control input u. Figure 6: The control input u with control scheme involv- ing sign() in [21] and with scheme in this paper. 0.15 12 z1 with control scheme in [21] z2 with control scheme in [21] z1 with our proposed shceme z2 with our proposed shceme 0.1 -5 0.05 -10 0 0.5 1 1.5 2 2.5 3 -2 -0.05 -4 -0.1 -6 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Time (seconds) Time (seconds) Figure 7: The output tracking error z with control Figure 8: The virtual error z with control scheme involv- 1 2 scheme involving sign() in [21] and with scheme in this ing sign() in [21] and with scheme in this paper. paper. 10 0.15 u with control scheme in [21] z1 with control scheme in [21] u with our proposed shceme z1 with our proposed shceme 0 0.1 -5 -10 0.05 -15 -20 0 -25 -30 -0.05 -35 -40 -0.1 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Time (seconds) Time (seconds) Figure 9: The control input u with control scheme involv- Figure 10: The output tracking error z with control ing arctan(10) in [21] and with scheme in this paper. scheme involving arctan(10) in [21] and with scheme in this paper. 10 12 40 z2 with control scheme in [21] z2 with our proposed shceme 10 20 8 0 6 -20 4 -40 -2 0 -4 -2 1.5 x 0.5 -6 2 -4 -0.5 0 10 20 30 40 50 60 70 80 90 100 -1 -1.5 Time (seconds) x Figure 11: The virtual error z with control scheme in- Figure 12: The state variables of uncontrolled system (38) volving arctan(10) in [21] and with scheme in this paper. with green dot and red dot representing initial condition and the origin respectively. 0.6 0 0.4 0.2 -10 -20 -0.2 2 -0.4 9 10 11 12 13 14 15 -2 -1 -4 -2 -6 0.8 0.7 0.6 0.5 0.4 2 0.3 -8 0.2 0.1 -3 0 10 20 30 40 50 60 70 80 90 100 -0.1 -0.2 Time (seconds) Figure 13: The state variables of (38) under proposed Figure 14: The virtual errors z and z . 2 3 control scheme with green dot and red dot representing initial condition and the origin respectively. -2 -4 -6 9 10 11 12 13 14 15 -20 -40 -60 0 10 20 30 40 50 60 70 80 90 100 Time (seconds) Figure 15: The control input u of (38). 5 Conclusion In this paper, we propose a smooth adaptive backstepping control design scheme for a class of SISO commensurate fractional-order nonlinear systems in strict feedback form with uncertain system parameters and unknown external 3 time-varying disturbance. It is proved that the resulting closed-loop system is globally asymptotically stable, even in the presence of arbitrary uncertainties and bounded disturbances. Simulation results also demonstrate the effectiveness in stabilizing unstable system and tracking reference signal with better performances compared to an existing scheme presented in [21]. For fractional order > 1, the control protocol in this paper is not theoretically shown effective, and therefore it is an interesting future research topic to extend our result to such systems. References [1] A Anatolii Aleksandrovich Kilbas, Hari Mohan Srivastava, and Juan J Trujillo. Theory and Applications of Fractional Differential Equations, volume 204. Amsterdam, Boston: Elsevier, 2006. [2] Igor Podlubny. Fractional-order systems and fractional-order controllers. Institute of Experimental Physics, Slovak Academy of Sciences, Kosice, 12(3):1–18, Nov. 1994. [3] Peter J Torvik and Ronald L Bagley. On the appearance of the fractional derivative in the behavior of real materials. Journal of Applied Mechanics, 51(2):294–298, Jun. 1984. [4] Michele Caputo and Francesco Mainardi. A new dissipation model based on memory mechanism. Pure and Applied Geophysics, 91(1):134–147, Dec. 1971. [5] CHR Friedrich. Relaxation and retardation functions of the Maxwell model with fractional derivatives. Rheologica Acta, 30(2):151–158, Mar. 1991. [6] Raoul R Nigmatullin and Stuart O Nelson. Recognition of the “fractional” kinetics in complex systems: dielectric properties of fresh fruits and vegetables from 0.01 to 1.8 GHz. Signal Processing, 86(10):2744–2759, Oct. 2006. [7] RC Koeller. Applications of fractional calculus to the theory of viscoelasticity. Journal of Applied Mechanics, 51(2):299–307, Jun. 1984. [8] Juan Yu, Cheng Hu, and Haijun Jiang. -stability and -synchronization for fractional-order neural networks. Neural Networks, 35:82–87, Nov. 2012. [9] JC Wang. Realizations of generalized Warburg impedance with RC ladder networks and transmission lines. Journal of the Electrochemical Society, 134(8):1915–1920, Aug. 1987. [10] Huiyang Liu, Long Cheng, Min Tan, and Zeng-Guang Hou. Exponential finite-time consensus of fractional-order multiagent systems. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 50(4):1549–1558, Apr. 2020. [11] Tom T Hartley and Carl F Lorenzo. Dynamics and control of initialized fractional-order systems. Nonlinear Dynamics, 29(1-4):201–233, Jul. 2002. [12] YangQuan Chen, Hyo-Sung Ahn, and Dingyü Xue. Robust controllability of interval fractional order linear time invariant systems. Signal Processing, 86(10):2794–2802, Jun. 2006. [13] YangQuan Chen, Ivo Petras, and Dingyu Xue. Fractional order control - A tutorial. In 2009 American Control Conference, pages 1397–1411, 2009. [14] Wei Wang, Changyun Wen, and Jing Zhou. Adaptive Backstepping Control of Uncertain Systems with Actuator Failures, Subsystem Interactions, and Nonsmooth Nonlinearities. Boca Raton: CRC Press, 2017. [15] Dumitru Baleanu, José António Tenreiro Machado, and Albert CJ Luo. Fractional Dynamics and Control. New York: Springer Science & Business Media, 2011. [16] Yiheng Wei, Yuquan Chen, Shu Liang, and Yong Wang. A novel algorithm on adaptive backstepping control of fractional order systems. Neurocomputing, 165:395–402, Oct. 2015. [17] Dian Sheng, Yiheng Wei, Songsong Cheng, and Yong Wang. Adaptive backstepping state feedback control for fractional order systems with input saturation. IFAC-PapersOnLine, 50(1):6996–7001, Jul. 2017. [18] Yiheng Wei, W Tse Peter, Zhao Yao, and Yong Wang. Adaptive backstepping output feedback control for a class of nonlinear fractional order systems. Nonlinear Dynamics, 86(2):1047–1056, Jul. 2016. [19] Shabnam Pashaei and Mohammadali Badamchizadeh. A new fractional-order sliding mode controller via a nonlinear disturbance observer for a class of dynamical systems with mismatched disturbances. ISA Transactions, 63:39–48, Jul. 2016. [20] Mou Chen, Shu-Yi Shao, Peng Shi, and Yan Shi. Disturbance-observer-based robust synchronization control for a class of fractional-order chaotic systems. IEEE Transactions on Circuits and Systems II: Express Briefs, 64(4):417–421, Apr. 2017. [21] Heng Liu, Yongping Pan, Shenggang Li, and Ye Chen. Adaptive fuzzy backstepping control of fractional-order nonlinear systems. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 47(8):2209–2217, Aug. 2017. 12 [22] Shuai Song, Baoyong Zhang, Jianwei Xia, and Zhengqiang Zhang. Adaptive backstepping hybrid fuzzy sliding mode control for uncertain fractional-order nonlinear systems based on finite-time scheme. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 50(4):1559–1569, Apr. 2020. [23] Zhen Wang. Synchronization of an uncertain fractional-order chaotic system via backstepping sliding mode control. Discrete Dynamics in Nature and Society, 2013, Jun. 2013. [24] Ping Gong and Weiyao Lan. Adaptive robust tracking control for multiple unknown fractional-order nonlinear systems. IEEE Transactions on Cybernetics, 49(4):1365–1376, Apr. 2019. [25] Qiao Wang, Jianliang Zhang, Dongsheng Ding, and Donglian Qi. Adaptive Mittag-Leffler stabilization of a class of fractional order uncertain nonlinear systems. Asian Journal of Control, 18(6):2343–2351, Nov. 2016. [26] Dongsheng Ding, Donglian Qi, and Qiao Wang. Non-linear Mittag–Leffler stabilisation of commensurate fractional-order non-linear systems. IET Control Theory & Applications, 9(5):681–690, Mar. 2015. [27] Dongsheng Ding, Donglian Qi, Yao Meng, and Li Xu. Adaptive Mittag-Leffler stabilization of commensurate fractional-order nonlinear systems. In 53rd IEEE Conference on Decision and Control, pages 6920–6926, 2014. [28] Zhiyao Ma and HongJun Ma. Adaptive fuzzy backstepping dynamic surface control of strict-feedback fractional order uncertain nonlinear systems. IEEE Transactions on Fuzzy Systems, 28(1):122–133, Jan. 2020. [29] Shuai Song, Baoyong Zhang, Xiaona Song, and Zhengqiang Zhang. Neuro-fuzzy-based adaptive dynamic surface control for fractional-order nonlinear strict-feedback systems with input constraint. IEEE Transactions on Systems, Man, and Cybernetics: Systems, to be published, doi: 10.1109/TSMC.2019.2933359. [30] Igor Podlubny. Fractional Differential Equations: An Introduction to Fractional Derivatives, Fractional Differen- tial Equations, to Methods of their Solution and some of their Applications, volume 198. San Diego: Academic Press, 1999. [31] Dingyü Xue. Fractional-Order Control Systems: Fundamentals and Numerical Implementations, volume 1. Berlin: De Gruyter, 2017. [32] V. Lakshmikantham, Srinivasa Leela, and J. Vasundhara Devi. Theory of Fractional Dynamic Systems. Cambridge: CSP, 2009. [33] Changpin Li and Weihua Deng. Remarks on fractional derivatives. Applied Mathematics and Computation, 187(2):777–784, Apr. 2007. [34] Bijnan Bandyopadhyay and Shyam Kamal. Stabilization and Control of Fractional Order Systems: A Sliding Mode Approach, volume 317. Cham: Springer, 2015. [35] Norelys Aguila-Camacho, Manuel A Duarte-Mermoud, and Javier A Gallegos. Lyapunov functions for fractional order systems. Communications in Nonlinear Science and Numerical Simulation, 19(9):2951–2957, Sep. 2014. [36] Yan Li, YangQuan Chen, and Igor Podlubny. Mittag–Leffler stability of fractional order nonlinear dynamic systems. Automatica, 45(8):1965–1969, Aug. 2009. [37] Yan Li, YangQuan Chen, and Igor Podlubny. Stability of fractional-order nonlinear dynamic systems: Lyapunov direct method and generalized Mittag–Leffler stability. Computers & Mathematics with Applications, 59(5):1810– 1821, Mar. 2010. [38] Fengrong Zhang, Changpin Li, and YangQuan Chen. Asymptotical stability of nonlinear fractional differential system with Caputo derivative. International Journal of Differential Equations, 2011, Aug. 2011. [39] Javier A Gallegos, Manuel A Duarte-Mermoud, Norelys Aguila-Camacho, and Rafael Castro-Linares. On fractional extensions of Barbalat lemma. Systems & Control Letters, 84:7–12, Oct. 2015. [40] Ruoxun Zhang and Yongli Liu. A new Barbalat’s lemma and Lyapunov stability theorem for fractional order systems. In 2017 29th Chinese control and decision conference (CCDC), pages 3676–3681, 2017. [41] Zongyu Zuo and Chenliang Wang. Adaptive trajectory tracking control of output constrained multi-rotors systems. IET Control Theory & Applications, 8(13):1163–1174, Sep. 2014. [42] Dong Yang, Guangdeng Zong, and Hamid Reza Karimi. H refined anti-disturbance control of switched LPV systems with application to aero-engine. IEEE Transactions on Industrial Electronics, 67(4):3180–3190, Apr. [43] Guangdeng Zong, Yankai Li, and Haibin Sun. Composite anti-disturbance resilient control for Markovian jump nonlinear systems with general uncertain transition rate. Science China Information Sciences, 62(2):22205, Jan. [44] Haibin Sun, Yankai Li, Guangdeng Zong, and Linlin Hou. Disturbance attenuation and rejection for stochastic Markovian jump system with partially known transition probabilities. Automatica, 89:349–357, Mar. 2018. 13 [45] Huaguang Zhang, Jian Han, Chaomin Luo, and Yingchun Wang. Fault-tolerant control of a nonlinear system based on generalized fuzzy hyperbolic model and adaptive disturbance observer. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 47(8):2289–2300, Aug. 2017. [46] Guy Jumarie. Modified Riemann-Liouville derivative and fractional Taylor series of nondifferentiable functions further results. Computers & Mathematics with Applications, 51(9-10):1367–1376, May 2006. [47] Jian Wang, Yanqing Wen, Yida Gou, Zhenyun Ye, and Hua Chen. Fractional-order gradient descent learning of BP neural networks with Caputo derivative. Neural Networks, 89:19–30, May 2017. [48] Ivo Petráš. A note on the fractional-order Chua’s system. Chaos, Solitons & Fractals, 38(1):140–147, Oct. 2008.

Journal

Electrical Engineering and Systems Science – arXiv (Cornell University)

Published: Feb 1, 2020

There are no references for this article.