A Gathering of Barbalat's Lemmas and Their (Unsung) Cousins: Keywords

A gathering of Barbalat’s lemmas and their
(unsung) cousins
Zhiyong Sun
arXiv:2301.00466v1 [math.OC] 1 Jan 2023
Control Systems Group, Department of EE, TU Eindhoven, the Netherlands
Abstract
This note presents a summary and review of various conditions for signal convergences,
based on Barbalat-like lemmas and their variations.
Keywords:
Signal convergence, function Lp space, Barbalat’s lemma.
1. Definitions and notations
Consider a function f (t) : R≥0 → R that is locallyR ∞ integrable. Given a fixed

p ∈ (0, ∞), we say that f (t) belongs to the Lp space if 0 |f (s)|p ds < ∞. We define
R∞ 1
the function p-norm as kf (t)kLp = ( 0 |f (s)|p ds) p , with a fixed p ∈ [1, ∞). The
L∞ norm for a function f (t) is defined as kf (t)k∞ = supt |f (t)|. We say a function
f (t) ∈ L∞ if supt |f (t)| < ∞.
For a vector-valued function f R(t) : R≥0 → Rn , we say that f (t) belongs to Lpn
∞
space (for a fixed p ∈ (0, ∞)) if 0 kf (s)kp ds < ∞ where k · k is the Euclidean
R∞ 1
norm1. We define the function p-norm in Rn as kf (t)kLpn = ( 0 kf (s)kp ds) p , with
a fixed p ∈ [1, ∞).
2. Barbalat’s lemma for signal convergence
The original version of the Barbalat’s lemma is stated as below.

Lemma 1. (Khalil, 2002) Consider a scalar uniformly continuous function f (t) such
Rt
that limt→∞ 0 f (τ )dτ exists and is finite. Then it holds limt→∞ f (t) = 0.
A sufficient condition to ensure that a function f (t) is uniformly continuous is that
f˙(t) is bounded. Therefore, the first variation of the Barbalat’s lemma is stated as
below.
Lemma 2. Consider a scalar function f (t) such that f (t) ∈ L1 and f˙(t) ∈ L∞ . Then
it holds limt→∞ f (t) = 0.
1 Note that any other vector norm could be used for the definition, due to equivalence of vector norms.
Preprint submitted to Nowhere January 3, 2023

3. Cousins of Barbalat’s lemma
Tao (Tao, 1997) proved the following simple alternative of Barbalat’s lemma.
Lemma 3. (Tao, 1997) If f (t) ∈ L2 and f˙(t) ∈ L∞ , then limt→∞ f (t) = 0.
Remark 1. Some remarks are in order.
• The condition “f˙(t) ∈ L∞ ” in the above lemma can be replaced by the more
general condition that “f (t) is uniformly continuous”; indeed, the condition
“f˙(t) ∈ L∞ ” is a sufficient condition to ensure that “f (t) is uniformly continu-
ous”.
• The proof of the above lemma also indicates that f (t) ∈ L∞ , and therefore
f (t) ∈ Lp ∩ L∞ , for any p ∈ [2, ∞).
• Rather than stating the condition “f (t) ∈ L2 ”, a common condition (in many
adaptive control books) is that “f (t) ∈ L2 ∩L∞ ”, but the boundedness condition
f (t) ∈ L∞ is not explicitly stated as it can be inferred by the two conditions on
f˙(t) and f (t) in the statement.
The following lemma is shown in (Desoer and Vidyasagar, 2009).
Lemma 4. (Desoer and Vidyasagar, 2009) Consider a scalar function f (t) such that
f (t) ∈ L2 and f˙(t) ∈ L2 . Then it holds that f (t) ∈ L∞ and limt→∞ f (t) = 0.
The following variation of Barbalat’s lemma is given in (Krstic et al., 1995).
Lemma 5. (Krstic et al., 1995, Page 491, Corollary A.7) Consider the function f (t) :
R≥0 → R. If f, f˙ ∈ L∞ , and f ∈ Lp for some p ∈ [1.∞), then limt→∞ f (t) = 0.
An alternative (and more general) lemma is proved in Tao’s book (Tao, 2003).
Lemma 6. (Tao, 2003, Page 80, Lemma 2.15) If f (t) ∈ Lp , 0 < p < ∞ and f˙(t) ∈
L∞ , then limt→∞ f (t) = 0.
Remark 2. Note that in the statement of Lemma 6, the condition on the function Lp
space is relaxed to include p ∈ (0, ∞). This is in contrast with the conditions in other
Barbalat-like lemmas, while often one imposes the condition p ∈ [1, ∞).
4. A more general version of Barbalat-like lemma
The following more general version of a Barbalat-like lemma is proved in (Farkas and Wegner,
2016).
Lemma 7. If f (t) ∈ Lp , p ∈ [1, ∞) and f˙(t) ∈ Lq , q ∈ (1, ∞], then f (t) is bounded
(i.e., f (t) ∈ L∞ ) and uniformly continuous. Furthermore, limt→∞ f (t) = 0.
• In the above lemmas, the conditions “f˙(t) ∈ L∞ ” or “f˙(t) ∈ Lq ” can be
replaced by that “f (t) is uniformly continuous”;
2
• The above lemmas for scalar function f (t) can be extended to vector-valued
functions f (t) : R≥0 → Rn .
• In Lemma 7, the condition f˙(t) ∈ Lq , q ∈ (1, ∞] can be generalized to that q ∈

[1, ∞] (i.e., the condition on q also includes the case that q = 1). Note: one can
prove that if f˙(t) ∈ L1 then limt→∞ f (t) exists and is finite. This fact together
with the condition f (t) ∈ Lp , p ∈ [1, ∞) will imply that limt→∞ f (t) = 0.
• The paper (Farkas and Wegner, 2016) also discussed the rate of convergence of
Lp functions. Examples are given in (Farkas and Wegner, 2016) to show the
relations of different versions of Barbalat-like lemmas.
• In the statement of Lemma 6, the condition on the function Lp space is relaxed
to include p ∈ (0, ∞). It is possible to further relax the conditions of Lemma 7
that incorporate the conditions of Lemma 6.
5. Other versions of Barbalat-like lemmas with function compositions
Lemma 8. (See e.g., (Teel, 1999)) Let α(t) : R≥0 → R≥0 be continuous, zero only at
zero, and nondecreasing. If f (t) is uniformly continuous on [0, ∞) and α(f (t)) ∈ L1 ,
then limt→∞ f (t) = 0.
A similar lemma is discussed in (Hou et al., 2010).
Lemma 9. Let M (z) be a continuous positive definite function defined on {z : z ∈
Rn , kzk ≤ r} for some r. If f (t) : R≥0 → Rn is a uniformly continuous function such
that f (t) ∈ L∞ 1
n and M (f (t)) ∈ L , then limt→∞ f (t) = 0.
• The convergence result for scalar functions f (t) in Lemma 8 can be extended to
vector-valued functions f (t) : R≥0 → Rn .
• By invoking the conditions of Lp functions in Lemma 6 and Lemma 7, it is pos-
sible to further relax the conditions of L1 (i.e., p = 1) function to other Lp
functions.
• The two lemmas are often applied (together with Lyapunov argument and com-
parison functions) to derive asymptotic convergence and Lp stability of dynami-
cal control systems.
3
References
Desoer, C.A., Vidyasagar, M., 2009. Feedback systems: input-output properties.

SIAM.
Farkas, B., Wegner, S.A., 2016. Variations on Barbălat’s lemma. The American Math-
ematical Monthly 123, 825–830.
Hou, M., Duan, G., Guo, M., 2010. New versions of Barbalat’s lemma with applica-
tions. Journal of Control Theory and Applications 8, 545–547.
Khalil, H.K., 2002. Nonlinear systems, Third Edition. Prentice hall New Jersey.
Krstic, M., Kokotovic, P.V., Kanellakopoulos, I., 1995. Nonlinear and adaptive control
design. John Wiley & Sons, Inc.
Tao, G., 1997. A simple alternative to the Barbalat lemma. IEEE Transactions on
Automatic Control 42, 698.
Tao, G., 2003. Adaptive control design and analysis. volume 37. John Wiley & Sons.
Teel, A.R., 1999. Asymptotic convergence from Lp stability. IEEE Transactions on
Automatic Control 44, 2169–2170.

A Gathering of Barbalat's Lemmas and Their (Unsung) Cousins: Keywords

Uploaded by

Copyright:

Available Formats

A Gathering of Barbalat's Lemmas and Their (Unsung) Cousins: Keywords

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Gathering of Barbalat's Lemmas and Their (Unsung) Cousins: Keywords

Uploaded by

Copyright:

Available Formats

A gathering of Barbalat’s lemmas and their

Control Systems Group, Department of EE, TU Eindhoven, the Netherlands

1. Definitions and notations

Consider a function f (t) : R≥0 → R that is locallyR ∞ integrable. Given a fixed

2. Barbalat’s lemma for signal convergence

The original version of the Barbalat’s lemma is stated as below.

Preprint submitted to Nowhere January 3, 2023

4. A more general version of Barbalat-like lemma

• In Lemma 7, the condition f˙(t) ∈ Lq , q ∈ (1, ∞] can be generalized to that q ∈

5. Other versions of Barbalat-like lemmas with function compositions

Remark 4. Some remarks are in order.

Desoer, C.A., Vidyasagar, M., 2009. Feedback systems: input-output properties.

You might also like