4 XIUYUN ZHENG · YUQIANG WU
Remark 1 It is worth pointing out that the transformation (7) has no definition at
u
0
= 0. In order to guarantee the feasibility of the coming control design and performance
analysis, without loss of generality, we assume that u
0
6= 0.
In order to remove the obstacle of zero crossings of u
0
due to the parameter update and
strong nonlinear drifts in the x
0
-subsystem, we design the control input u
0
such that it dom-
inates the nonlinear drifts and vanishes only when the state x
0
approaches to the origin. We
consider two cases, x
0
(t
0
) = 0 and x
0
(t
0
) 6= 0. Let us consider x
0
(t
0
) 6= 0 first, the case
x
0
(t
0
) = 0 will be discussed in section 4. Take u
0
as follows
u
0
= −c
0
x
0
− x
0
¯
f
0
(x
0
), c
0
> 0, (9)
where
¯
f
0
(x
0
) are given in (2), c
0
is a p ositive design parameter. Consequently, for the first
equation of (1), the following lemma can be established.
Lemma 1 For any initial condition (t
0
, x
0
(t
0
)), where t
0
≥ 0, x
0
(t
0
) ∈ R
1
, if u
0
is given in
(9), then the corresponding solution x
0
(t) exists and satisfies |x
0
(t, t
0
, x
0
(t
0
))| > 0 for t ≥ t
0
and x
0
(t
0
) 6= 0.
Proof Substituting (9) into the first equation of (1), we have
˙x
0
= −c
0
x
0
− x
0
¯
f
0
(x
0
) + f
0
(x
0
). (10)
Introduce the Lyapunov function V
0
(x
0
) =
1
2
x
2
0
, choosing the parameter c
0
> 1 and using the
condition (2), we obtain that the time derivative of V
0
(x
0
) along the solution of (10) satisfies
˙
V
0
= x
0
(−c
0
x
0
− x
0
¯
f
0
(x
0
) + f
0
(x
0
)) ≤ −c
0
x
2
0
. (11)
This implies that x
0
(t) tends to zero exponentially.
In the following, we will conclude that x
0
(t) does not become zero in any time instant. In
fact for any bounded x
1
(t) and x
0
(t), since
¯
f
0
(x
0
) are smooth function with respect to their
arguments, there exists a positive constant M
1
< 1 such that
¯
f
0
(x
0
) ≤ M
1
< 1 for |x
0
| <
¯
δ
1
≤ 1.
This guarantees that as |x
0
| <
¯
δ
1
≤ 1, the following inequality
˙
V
0
≥ −ρx
2
0
, (12)
holds, where ρ = M
1
+ Mc
0
, M = max(
¯
f
0
(x
0
)) + 1. This implies that |x
0
(t)| converges to zero
with a rate less than a certain constant ρ. This completes the proof of the lemma.
If x
0
(t
0
) 6= 0. Due to the existence of the nonlinear drifts in the dynamical equation of
x
0
, a linear feedback control with constant feedback in the form of u
0
= −k
0
x
0
(where k
0
is
a constant) is not adequate. The reason is that the constant k
0
may not be large enough to
dominate the uncontrolled nonlinear terms, as will be clear later. Therefore, in this paper, we
will design the control in the form
u
0
= −πx
0
, x
0
(t
0
) 6= 0, (13)
where π is a smooth feedback function.
Lemma 2 Given any C
1
function g(x
1
, · · · , x
n
), there exist some continuous functions
¯g
i
(x
1
, · · · , x
n
), i = 1, 2, · · · , n such that
g(x
1
, · · · , x
n
) = g(0, 0 · · · , 0) +
n
X
i=1
x
i
¯g
i
(x
1
, · · · , x
n
). (14)
According to the first equation of (1), Lemma 2 and the definition (9) of u
0
, we have
˙u
0
u
0
= β +
¯
φ
0
(x
0
), (15)
http://www.paper.edu.cn