Professional Documents
Culture Documents
Discussion On Barbalat Lemma Extensions For Conformable Fractional Integrals
Discussion On Barbalat Lemma Extensions For Conformable Fractional Integrals
To cite this article: A. Souahi, O. Naifar, A. Ben Makhlouf & M. A. Hammami (2017): Discussion
on Barbalat Lemma extensions for conformable fractional integrals, International Journal of Control,
DOI: 10.1080/00207179.2017.1350754
Download by: [Cornell University Library] Date: 14 July 2017, At: 15:37
INTERNATIONAL JOURNAL OF CONTROL,
https://doi.org/./..
KEYWORDS
Conformable fractional
derivative; Barbalat-like
lemmas; conformable
fractional differential
equation
Some variants and generalisations were derived for the In addition, if the conformable fractional derivative of
Barbalat lemma, an interesting one appeared in the lit- f of order α exists on (a, ) , then we say that f is α-
erature of control theory see, Tao (1997). If differentiable on (a, ).
Lemma 2.1 (Eslami & Rezazadeh, 2015): Let α (0, 1),
+∞ c1 , c2 , r, λ ∈ R and the functions f, g : [a, +∞) −→ R be
e2 (s)ds < ∞ α-differentiable on (a, ). Then
0
Example 3.1: We would like to find a function f which is Since the term l−ε (t α − T α ) tends to infinity when t →
α
nonnegative, uniformly continuous, and Iα f is bounded + we obtain a contradiction.
but f does not converge to zero.
Let P be a function null at every point except the inter- Now, for the case of Barbalat-type lemma proposed
vals [ui , ui + δ] in which it takes the value 1, where δ = 12 in Tao (1997), we prove the following conformable ana-
and ui = iγ , with γ = 1−α 2
. logue.
We have, Theorem 3.1: Let f : R+ −→ R p be a function differen-
tiable on R∗+ .
+∞
ui +δ
α If there exists some α (0, 1), such that
I P(t ) ≤ sα−1 ds < ∞.
i=0 ui
+∞
sα−1 f (s) γ ds < ∞, where γ > 0, (4)
Now, let 0
⎧
⎨ t − ui , if t ∈ [ui , ui + 2δ ] and Tα f is a bounded function, then,
f (t ) = ui + δ − t, if t ∈ [ui + 2δ , ui + δ]
⎩
0, if not lim f (t ) = 0.
t−→+∞
We have f P, then Iα f is bounded. Proof: Since Tα f is a bounded function. Thus, there exists
Hence, f is nonnegative, uniformly continuous func- M > 0 such that s1 − α f (s) M, s > 0.
tion and Iα f is bounded but f does not converge to zero. Suppose that, limt−→+∞ f (t ) = 0. So, there exists a
sequence tn −→ +∞ such that f(tn ) δ,∀n ∈ N∗ , for
Example 3.1 shows that there is no analogue of such
some δ > 0. Without loss of generality, we may assume
Barbalat lemma for conformable fractional integrals.
that t1α > 2M(2δαα −1) and tn+1 > 2tn , ∀n ∈ N∗ , by replacing
We introduce another result related to the same Bar-
(tn )n∈N∗ by a suitable subsequence.
balat lemma (case Khalil, 1996; Popov, 1973; Slotine & Li, δα α1
For each n ∈ N∗ , one chooses sn = [tnα + 2M ] . Then,
1991) for conformable fractional integrals. We have the
following result.
δα 1
Proposition 3.1: Let f : R+ −→ R p be a bounded func- tn ≤ sn = [tnα + ] α ≤ 2tn < tn+1 , ∀n ∈ N∗ .
2M
tion such that
By the generalised mean-value theorem,
+∞
α−1
s f (s) ds < ∞,
0 f (t ) − f (tn ) ≤ M(sn − tn )tnα−1 , ∀t ∈ [tn , sn ].
then, We have
lim inf f (t ) = 0.
1
α δα α
− (tnα ) α
1
t→+∞
s n − tn = t n +
2M
Proof: f is bounded, then lim inf f (t ) exists. Let tnα +δ α
t→+∞ 1
1 tn +δ 1 −1 δ s1−α
= [s ] ds ≤
α s α ds ≤ n ,
tnα α tnα α
l = lim inf f (t ) ≥ 0. (5)
t→+∞
|k|
T β X (t )T PX (t ) + y(t )T y(t )
V (e, y) = e2 + y2 . γ
|k|
Its β derivative is = 2X (t )T PT β X (t ) + 2 y(t )T T β y(t )
γ
= 2X (t ) P AX (t ) + kby(t )T u(t )
T
T β V (t ) = 2e(t ) − e(t ) + y(t )w(t ) + 2y(t )(−ew(t ))
|k|
= −2e2 (t ) ≤ 0. + 2 y(t )T − γ sgn(k)X T (t )Pbu(t )
γ
INTERNATIONAL JOURNAL OF CONTROL 5
0.8
e y
0.6
0.4
Solutions
0.2
−0.2
−0.4
0 5 10 15
Time (s)
= X (t )T AT P + PA X (t ) It follows that,
= −X (t ) QX (t ) T
+∞
= −λmin (Q) X (t ) . 2
(13) sβ−1 X (s) 2 ds < ∞.
0
By (13), we have Since u(t) is bounded, then by Theorem 3.1, X(t) con-
verges to zero as t −→ +∞.
|k|
T β X (t )T PX (t ) + y(t )T y(t ) ≤ 0. For this example, we choose
γ
Thus, we obtain −5 1
A= ,
0 −5
|k|
X (t )T PX (t ) + y(t )T y(t ) 2
γ b= ,
1
|k|
≤ X (0)T PX (0) + y(0)T y(0) . (14)
γ u(t) = e−t , k = 2, γ = 1 and β = 0.5. Using the MATLAB
LMI control toolbox, the matrix P is given by
Hence, X(t) and y(t) are bounded t 0.
By (13), we get
0.1356 0.0145
P= .
|k| 0.0145 0.1383
β
λmin (Q) X (t ) ≤ −T X (t ) PX (t ) +
T2
y(t ) y(t ) .
T
γ
(15) The initial conditions are defined as follows: X1 (0) = −2,
X2 (0) = 1 and y(0) = 1. The plot of the solutions X(t) =
Hence, (X1 (t), X2 (t)) and y(t) are shown in Figures 2 and 3.
We would like to mention that the case of β = 1 of
t
this example was already taken and solved in Narendra
λmin (Q) sβ−1 X (s) 2 ds
and Annaswamy (2005). The Caputo fractional derivative
0 case, where β (0, 1), was presented in Aguila-Camacho
|k|
≤ X (0)T PX (0) + y(0)T y(0) and Duarte-Mermoud (2016). It should be underlined
γ
that in this case, that is: Caputo definition, solutions con-
|k|
− X (t ) PX (t ) − y(t ) y(t ) .
T T
(16) verge in mean value to zero and not towards zero.
γ
6 A. SOUAHI ET AL.
x x
1 2
0.5
0
Solutions (X)
−0.5
−1
−1.5
−2
0 1 2 3 4 5 6 7 8 9 10
Time (s)
1.05
0.95
0.9
0.85
y
0.8
0.75
0.7
0.65
0 5 10 15
Time (s)
As an interesting results presented in this work, the For the first case of Barbalat-like lemmas, it was proved
solutions of such systems, using conformable fractional via an illustrative counterexample that there is no
integrals, converge directly to zero which is well validated straightforward extension. For the second generalised
throughout these examples. Barbalat lemma, an analogue for conformable fractional
integrals is provided. For applications in control theory
(such that adaptive control, adaptive observer and so
5. Conclusion on), the second extension of Barbalat lemma fulfills the
Some extensions for the conformable fractional required condition and guarantees the convergence of
order case of the well-known Barbalat lemma for the the solution towards zero. This is an interesting result
integer case have been discussed throughout the paper. since it will be very useful especially for adaptive systems.
INTERNATIONAL JOURNAL OF CONTROL 7
Naifar, O., Ben Makhlouf, A., Hammami, M. A., & Ouali, A. Tasbozan, O., Cenesiz, Y., & Kurt, A. (2016). New solutions
(2015). State feedback control law for a class of nonlinear for conformable fractional Boussinesq and combined KdV–
time–varying system under unknown time-varying delay. mKdV equations using Jacobi elliptic function expansion
Nonlinear Dynamics, 82, 349–355. method. The European Physical Journal Plus (pp. 131–244).
Narendra, K. S., & Annaswamy, M. (2005). Stable adaptive sys- doi:10.1140/epjp/i2016-16244-x
tems. New York, NY: Dover Publications Inc. Wang, L. L., & Fu, J. L. (2016). Non-Noether symmetries of
Popov, V. M. (1973). Hyperstability of control systems. New York, Hamiltonian systems with conformable fractional deriva-
NY: Springer-Verlag. tives. Chinese Physics B, 25, 014501.
Slotine, J. J. E., & Li, W. (1991). Applied of nonlinear control. Yu, X., & Wu, Z. (2014). Corrections to ‘Stochastic Barbalats
Englewood Cliffs, NJ: Prentice-Hall. lemma and its applications”. IEEE Transactions on Auto-
Tao, G. (1997). A simple alternative to the Barbalat lemma. IEEE matic Control, 59, 1386–1390.
Transactions on Automatic Control, 42(5), 698.