In what follows we assume that the following standing hypothesis holds
Assumption 1
There exists a closed covering
$\chi $
of
${\mathbb{R}}^{n}$
such that
${\mathcal{T}}^{\prime}\subseteq \mathcal{T}\left[\chi \right]$
.
$\square $
Since we are mainly interested in the stability analysis of the zero solutions of system ( 1 ), we will also suppose the following hypothesis holds
Assumption 2
$f(0,\gamma )=0$
for all
$\gamma \in {\Gamma}^{*}$
, where
${\Gamma}^{*}=\{\gamma \in \Gamma :0\in {\chi}_{\gamma}\}$
;
$\square $
and adopt the following definitions of stability.
Definition 2.1
We say that a family
${\mathcal{T}}^{\prime}$
of maximal trajectories of ( 1 ) is

1.
uniformly stable if there exists a function
$\alpha \in {\mathcal{K}}_{\infty}$
such that for every
$(x,\sigma )\in {\mathcal{T}}^{\prime}$
,
$$\begin{array}{ccc}\leftx\right(t\left)\right\le \alpha \left(\rightx\left({t}_{0}\right)\left\right)\forall t\ge {t}_{0},\forall {t}_{0}\ge 0.& & \end{array}$$  

2.
Globally asymptotically stable if it is uniformly stable and, in addition, for every trajectory
$(x,\sigma )\in {\mathcal{T}}^{\prime}$
,
$x\left(t\right)$
converges to
$0$
as
$t\to \infty $
.

3.
Globally uniformly asymptotically stable if there exists a function
$\beta \in \mathcal{K}\mathcal{\mathcal{L}}$
such that for every
$(x,\sigma )\in {\mathcal{T}}^{\prime}$
,
$$\begin{array}{ccc}\leftx\right(t\left)\right\le \beta \left(\rightx\left({t}_{0}\right),t{t}_{0})\forall t\ge 0,\forall {t}_{0}\ge 0.& & \end{array}$$  
$\square $
Remark 2.1
With the same technique used to prove Proposition 2.5 of [
11]
, it can be shown that the definition of global uniform asymptotic stability given above is equivalent to the following (more classical) one:
A family
${\mathcal{T}}^{\prime}$
of maximal trajectories of ( 1 ) is globally uniformly asymptotically stable if it is uniformly stable and

(a)
for each
$R>0$
and each
$\varepsilon >0$
, there exists
$T\ge 0$
such that for all
$(x,\sigma )\in {\mathcal{T}}^{\prime}$
and all
${t}_{0}\ge 0$
$$\leftx\right({t}_{0}\left)\right<R\u27f9\leftx\right(t\left)\right<\varepsilon \forall t\ge {t}_{0}+T.$$
$\square $
Several Lyapunovlike theorems which involve the use of multiple Lyapunov functions (see, among others, [
2]
, [
3]
) allow us to establish the stability or asymptotic stability of a family of admissible trajectories. The following one which is based on results given in [
2]
, [
13]
(see also [
5]
) is an example of such theorems. It is convenient to introduce here the following
Definition 2.2
A function
$V:{\mathbb{R}}^{n}\times \Gamma \to \mathbb{R}$
is a weak Lyapunovlike function for the family
${\mathcal{T}}^{\prime}$
if it is continuously differentiable with respect to the first argument and verifies

1.
there exist
${\alpha}_{1}$
and
${\alpha}_{2}$
of class
${\mathcal{K}}_{\infty}$
so that
${\alpha}_{1}\left(\right\xi \left\right)\le V(\xi ,\gamma )\le {\alpha}_{2}\left(\right\xi \left\right)$
for all
$(\xi ,\gamma )\in {\mathbb{R}}^{n}\times \Gamma $
such that
$\xi \in {\chi}_{\gamma}$
;

2.
$\frac{\partial V}{\partial \xi}(\xi ,\gamma )f(\xi ,\gamma )\le 0$
, for all
$(\xi ,\gamma )\in {\mathbb{R}}^{n}\times \Gamma $
such that
$\xi \in {\chi}_{\gamma}$
;

3.
for every trajectory
$(x,\sigma )\in {\mathcal{T}}^{\prime}$
and any pair
${t}_{i}<{t}_{j}$
of switching times such that
$\sigma \left({t}_{i}\right)=\sigma \left({t}_{j}\right)$
,
$V\left(x\right({t}_{j}),\sigma ({t}_{j}\left)\right)\le V\left(x\right({t}_{i+1}),\sigma ({t}_{i}\left)\right)$
.
$\square $
Theorem 2.1
Suppose there exists a weak Lyapunovlike function
$V$
for
${\mathcal{T}}^{\prime}$
. Then
${\mathcal{T}}^{\prime}$
is uniformly stable.
If, in addition,
$V$
verifies

2
${}^{\prime}$
.
there exists a positivedefinite function
${\alpha}_{3}$
such that
$\frac{\partial V}{\partial \xi}(\xi ,\gamma )f(\xi ,\gamma )\le {\alpha}_{3}\left(\right\xi \left\right)$
, for all
$(\xi ,\gamma )\in {\mathbb{R}}^{n}\times \Gamma $
such that
$\xi \in {\chi}_{\gamma}$
,
then
${\mathcal{T}}^{\prime}$
is globally uniformly asymptotically stable.
$\square $
As was pointed out above, this work is concerned with invariance principles for switched systems and, in particular, with extensions of LaSalle's invariance principle to this class of systems. In this regard we will show, under suitable hypotheses, that the existence of a weak Lyapunovlike function
$V$
for
${\mathcal{T}}^{\prime}$
, allows us to obtain conclusions about the asymptotic behavior of a bounded solution
$x$
of ( 1 ) corresponding to some switching signal
$\sigma $
so that
$(x,\sigma )\in {\mathcal{T}}^{\prime}$
, and, further, to obtain some asymptotic stability criteria. As was discussed in [5] and also in [1] , in order to obtain LaSallelike asymptotic stability criteria by exploiting the knowledge of a weak Lyapunovlike function
$V$
, some form of regularity in the switching signals regarding the distance between consecutive switching times is needed. In this paper we will consider switching signals which have a positive average dwelltime, more precisely,
Definition 2.3
We say that the switching signal
$\sigma $
has an average dwelltime
${\tau}_{D}>0$
and a chatter bound
${N}_{0}\in \mathbb{N}$
if the number of switching times of
$\sigma $
in any open finite interval
$({\tau}_{1},{\tau}_{2})\subset {\mathbb{R}}_{+}$
is bounded by
${N}_{0}+({\tau}_{2}{\tau}_{1})/{\tau}_{D}$
.
$\square $
We denote by
$\mathcal{S}a[{\tau}_{D},{N}_{0}]$
the set of all the switching signals which have an average dwelltime
${\tau}_{D}>0$
and a chatter bound
${N}_{0}\in \mathbb{N}$
and by
$\mathcal{T}a[{\tau}_{D},{N}_{0}]$
the subclass of all the trajectories of ( 1 ) corresponding to some
$\sigma \in \mathcal{S}a[{\tau}_{D},{N}_{0}]$
. Let
$\mathcal{S}a={\cup}_{{\tau}_{D}>0,{N}_{0}>0}\mathcal{S}a[{\tau}_{D},{N}_{0}]$
and let
$\mathcal{T}a$
denote the subclass of all the trajectories of ( 1 ) corresponding to some
$\sigma \in \mathcal{S}a$
, i.e.
$\mathcal{T}a={\cup}_{{\tau}_{D}>0,{N}_{0}>0}\mathcal{T}a[{\tau}_{D},{N}_{0}]$
.
We note that the set of switching signals
$\sigma $
which have a dwelltime
${\tau}_{D}>0$
, i.e.,
${inf}_{k\ge 0}{t}_{k+1}{t}_{k}\ge {\tau}_{D}$
, is a subset of
$\mathcal{S}[{\tau}_{D},1]$
.
From now on we suppose that the following additional hypothesis holds.
Assumption 3
${\mathcal{T}}^{\prime}\subseteq \mathcal{T}a$
.
$\square $
In order to establish the main results of this paper, we need to introduce some more trajectory families.
Given a continuous function
$V:\Omega \times \Gamma \to \mathbb{R}$
, with
$\Omega $
an open subset of
${\mathbb{R}}^{n}$
, we consider the following families of trajectories associated with
$V$
.
${\mathcal{T}}_{V}$
is the class of all the trajectories
$(x,\sigma )\in \mathcal{T}$
which verify the conditions:

1.
$x\left(t\right)\in \Omega $
for all
$t\in {\mathcal{\mathcal{I}}}_{(x,\sigma )}$
;

2.
for any pair of times
$t,{t}^{\prime}\in {\mathcal{\mathcal{I}}}_{(x,\sigma )}$
such that
$t\le {t}^{\prime}$
and
$\sigma \left(t\right)=\sigma \left({t}^{\prime}\right)$
,
$V\left(x\right(t),\sigma (t\left)\right)\ge V\left(x\right({t}^{\prime}),\sigma ({t}^{\prime}\left)\right)$
.
${\mathcal{T}}_{V}^{*}$
is the subfamily of
${\mathcal{T}}_{V}$
whose members
$(x,\sigma )$
verify the condition
$$\sigma \left(t\right)=\sigma \left({t}^{\prime}\right)\u27f9V\left(x\right(t),\sigma (t\left)\right)=V\left(x\right({t}^{\prime}),\sigma ({t}^{\prime}\left)\right).$$
Remark 2.2
If
$V$
is a weak Lyapunovlike function for
${\mathcal{T}}^{\prime}$
it readily follows that
${\mathcal{T}}^{\prime}\subseteq {\mathcal{T}}_{V}$
.
$\square $
Finally, we introduce the following notion of weakinvariance for nonempty subsets of
${\mathbb{R}}^{n}\times \Gamma $
.
Definition 2.4
Given a family
${\mathcal{T}}^{*}$
of maximal trajectories of ( 1 ), we say that a nonempty subset
$M\subseteq {\mathbb{R}}^{n}\times \Gamma $
is weaklyinvariant with respect to
${\mathcal{T}}^{*}$
if for each
$(\xi ,\gamma )\in M$
there is a trajectory
$(x,\sigma )\in {\mathcal{T}}^{*}$
such that
$x\left(0\right)=\xi $
,
$\sigma \left(0\right)=\gamma $
and
$\left(x\right(t),\sigma (t\left)\right)\in M$
for all
$t\in {\mathcal{\mathcal{I}}}_{(x,\sigma )}$
.
$\square $
Now we are in position to state the following asymptotic stability criterion, which is one of our main results.
Theorem 2.2
Suppose that there exists a weak Lyapunovlike function
$V$
for
${\mathcal{T}}^{\prime}$
such that
$M=\left\{0\right\}\times {\Gamma}^{*}$
is the maximal weaklyinvariant set w.r.t.
${\mathcal{T}}_{V}^{*}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a$
.
Then
${\mathcal{T}}^{\prime}$
is globally asymptotically stable. If, in addition,
${\mathcal{T}}^{\prime}\subseteq \mathcal{T}a[{\tau}_{D},{N}_{0}]$
for some
${\tau}_{D}>0$
and some
${N}_{0}\in \mathbb{N}$
then
${\mathcal{T}}^{\prime}$
is globally uniformly asymptotically stable.
$\square $
Remark 2.3
From the proof of Theorem 2.2 , which can be found in section 5, it follows that in the case when
${\mathcal{T}}^{\prime}\subseteq \mathcal{T}a[{\tau}_{D},{N}_{0}]$
for some
${\tau}_{D}>0$
and some
${N}_{0}\in \mathbb{N}$
, the thesis of the theorem still holds if one assumes the weaker hypothesis
$M=\left\{0\right\}\times {\Gamma}^{*}$
is the maximal weaklyinvariant set w.r.t.
${\mathcal{T}}_{V}^{*}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a[{\tau}_{D},{N}_{0}]$
.
$\square $
The following result can be readily deduced from Theorem 2.2 .
Theorem 2.3
Suppose that there exists a weak Lyapunovlike function
$V$
for
${\mathcal{T}}^{\prime}$
.
Suppose, in addition, that there exists a family
$\{{W}_{\gamma}:{\mathbb{R}}^{n}\to \mathbb{R},\gamma \in \Gamma \}$
of continuous and nonnegative definite functions such that

1.
$\frac{\partial V}{\partial \xi}(\xi ,\gamma )f(\xi ,\gamma )\le {W}_{\gamma}\left(\xi \right)$
, for all
$(\xi ,\gamma )\in {\mathbb{R}}^{n}\times \Gamma $
such that
$\xi \in {\chi}_{\gamma}$
;

2.
for each
$\gamma \in \Gamma $
, the system
$$\begin{array}{ccc}\dot{x}& =& {f}_{\gamma}\left(x\right),y={W}_{\gamma}\left(x\right),\end{array}$$ 
(2)

is zero smalltime distinguishable. (We recall that a systems ( 2 ) is zero smalltime distinguishable, if for every
$\delta >0$
,
$x\left(0\right)=0$
whenever
${W}_{\gamma}\left(x\right(t\left)\right)=0$
for all
$t\in [0,\delta )$
).
Then
${\mathcal{T}}^{\prime}$
is globally asymptotically stable. If, in addition,
${\mathcal{T}}^{\prime}\subseteq \mathcal{T}a[{\tau}_{D},{N}_{0}]$
for some
${\tau}_{D}>0$
and some
${N}_{0}\in \mathbb{N}$
then
${\mathcal{T}}^{\prime}$
is globally uniformly asymptotically stable.
$\square $
Proof. Let
$M$
be the maximal weaklyinvariant set w.r.t.
${\mathcal{T}}_{V}^{*}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a$
. In order to prove the theorem, it suffices to show that
$M=\left\{0\right\}\times {\Gamma}^{*}$
, since then the hypotheses of Theorem 2.2 will be fulfilled. Note first that
$\left\{0\right\}\times {\Gamma}^{*}$
is weaklyinvariant w.r.t.
${\mathcal{T}}_{V}^{*}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a$
due to Assumption 2 and the fact that, due to 1. in Definition 2.2 ,
$V(0,\gamma )=0\forall \gamma \in {\Gamma}^{*}$
.
Let then
$(\xi ,\gamma )\in M$
; it follows that there exists a trajectory
$(x,\sigma )\in {\mathcal{T}}_{V}^{*}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a$
such that
$\left(x\right(0),\sigma (0\left)\right)=(\xi ,\gamma )$
and
$\left(x\right(t),\sigma (t\left)\right)\in M$
for all
$t\ge 0$
. Let
$\tau >0$
such that
$\sigma \left(t\right)=\gamma $
for all
$t\in [0,\tau )$
.
It follows from the definition of
${\mathcal{T}}_{V}^{*}$
that
$\frac{\partial V}{\partial \xi}\left(x\right(t),\sigma (t\left)\right)f\left(x\right(t),\sigma (t\left)\right)=0$
for all
$t\in [0,\tau )$
and hence
${W}_{\gamma}\left(x\right(t\left)\right)=0$
for every
$t\in [0,\tau )$
. From the zero smalltime distinguishability assumption, we have that
$\xi =x\left(0\right)=0$
. Since
$(x,\sigma )\in \mathcal{T}\left[\chi \right]$
,
$x\left(0\right)\in {\chi}_{\sigma \left(0\right)}$
and consequently
$\gamma \in {\Gamma}^{*}$
.
Remark 2.4
Theorem 2.3 is a partial generalization of Theorem 7 in [
6]
, since the zero smalltime distinguishability hypothesis is weaker than the smalltime normobservability Assumption 2. in [
6]
and since we obtain uniform asymptotic stability in the case when
${\mathcal{T}}^{\prime}\subseteq \mathcal{T}a[{\tau}_{D},{N}_{0}]$
for some
${\tau}_{D}>0$
and some
${N}_{0}\in {\mathbb{N}}_{0}$
, but our hypothesis
${\mathcal{T}}^{\prime}\subseteq \mathcal{T}a$
is slightly stronger than the hypothesis about the regularity of the switching signals considered in [
6]
(see Assumption 3. of that paper).
$\square $
The proof of Theorem 2.2 is based on the following extension of the well known invariance principle for dynamical systems described by differential equations of LaSalle (see [
8]
) to switched systems. This extension is other of the main results of this work.
Let
${\pi}_{1}:{\mathbb{R}}^{n}\times \Gamma \to {\mathbb{R}}^{n}$
be the projection onto the first component.
Theorem 2.4
Let
$\chi $
be a closed covering of
${\mathbb{R}}^{n}$
and let
$V:\Omega \times \Gamma \to \mathbb{R}$
, with
$\Omega $
an open subset of
${\mathbb{R}}^{n}$
, be continuous. Suppose that
$(x,\sigma )$
is a trajectory belonging to
${\mathcal{T}}_{V}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a[{\tau}_{D},{N}_{0}]$
such that for some compact subset
$B\subset \Omega $
,
$x\left(t\right)\in B$
for all
$t\ge 0$
. Let
$M\subseteq {\mathbb{R}}^{n}\times \Gamma $
be the largest weaklyinvariant set w.r.t.
${\mathcal{T}}_{V}^{*}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a[{\tau}_{D},{N}_{0}]$
contained in
$\Omega \times \Gamma $
.
Then,
$x\left(t\right)$
converges to
${\pi}_{1}\left(M\right)$
as
$t\to \infty $
.
$\square $
Remark 2.5
In the case when we restrict the hypotheses of Theorem 2.4 to those of Theorems 1 and 2 in [
1]
, we obtain more precise results related to the size of the attracting sets involved. We shall prove our assertion for Theorem 1 in [
1]
only, since the proof for the other is similar. In what follows, and in order to prove our claim, we refer to the notation and definitions of that paper. Let
$x\left(t\right)$
be a dwelltime solution (in the sense of [
1]
) with initial condition
$x\left(0\right)\in {\Omega}_{l}$
, a dwelltime
${\tau}_{D}>0$
and generated by a switching signal
$\sigma $
in a switched system that admits a common weak Lyapunov function (in the sense above)
$V:\Omega \to {\mathbb{R}}_{+}$
.
Let
$W:{\Omega}_{l}\to {\mathbb{R}}_{+}$
be the restriction to
${\Omega}_{l}$
of the function
$V$
. It is easy to see that the trajectory
$(x,\sigma )$
belongs to
${\mathcal{T}}_{W}\cap \mathcal{T}a[{\tau}_{D},1]$
, and that there exists a compact set
$B\subset {\Omega}_{l}$
so that
$x\left(t\right)\in B$
for all
$t\ge 0$
.
Therefore the trajectory
$(x,\sigma )$
verifies the hypotheses of Theorem 2.4 (with
$W$
in place of
$V$
and the trivial covering of
${\mathbb{R}}^{n}$
,
${\chi}_{\gamma}={\mathbb{R}}^{n}$
for all
$\gamma \in \Gamma $
). Hence
$x\left(t\right)$
converges to
${\pi}_{1}\left(M\right)$
as
$t\to \infty $
, where
$M$
is the largest weaklyinvariant set w.r.t.
${\mathcal{T}}_{W}^{*}\cap \mathcal{T}a$
contained in
${\Omega}_{l}\times \Gamma $
.
On the other hand, Theorem 1 asserts that
$x\left(t\right)$
converges to
${M}^{\prime}$
, where
${M}^{\prime}$
is the union of all the compact, weaklyinvariant sets (in the sense of [
1]
) which are contained in
$Z\cap {\Omega}_{l}$
. We will prove that
${\pi}_{1}\left(M\right)\subseteq {M}^{\prime}$
and therefore our assertion about the sizes of the attracting sets. Pick
$\xi \in {\pi}_{1}\left(M\right)$
. Then there exist
$\gamma \in \Gamma $
and a trajectory
$({x}^{*},{\sigma}^{*})\in {\mathcal{T}}_{W}^{*}\cap \mathcal{T}a$
such that
$\left({x}^{*}\right(0),{\sigma}^{*}(0\left)\right)=(\xi ,\gamma )$
and
$\left({x}^{*}\right(t),{\sigma}^{*}(t\left)\right)\in M$
for all
$t\in {\mathcal{\mathcal{I}}}_{({x}^{*},{\sigma}^{*})}$
. In consequence
${\sigma}^{*}\left(t\right)=\gamma $
for all
$t\in [0,\delta ]$
with
$0<\delta <{t}_{1}$
and
${t}_{1}$
the first switching time of
${\sigma}^{*}$
. It follows from the definition of
${\mathcal{T}}_{W}^{*}$
that
$\frac{\partial V}{\partial \xi}\left({x}^{*}\right(t\left)\right)f\left({x}^{*}\right(t),\gamma )=0$
and hence that
${x}^{*}\left(t\right)\in Z\cap {\Omega}_{l}$
for all
$t\in [0,\delta ]$
. Next,
${x}^{*}\left(\right[0,\delta \left]\right)$
is a compact weaklyinvariant set contained in
$Z\cap {\Omega}_{l}$
and consequently
$\xi \in {M}^{\prime}$
.
It must be remarked that the attracting set
${\pi}_{1}\left(M\right)$
corresponding to the weaklyinvariant set considered in Theorem 2.4 , may be considerably smaller than the attracting sets given in Theorems 1 and 2 of [
1]
, as we exhibit in Example 2 below.
$\square $
3 Examples
Example 1. Consider the switched system in
${\mathbb{R}}^{2}$
given by the family
$\{{f}_{1},{f}_{2}\}$
, with
$$\begin{array}{ccc}{f}_{1}\left(\xi \right)=\left(\begin{array}{c}2{\xi}_{1}2{\xi}_{2}\\ 2{\xi}_{1}\end{array}\right),{f}_{2}\left(\xi \right)=\left(\begin{array}{c}{\xi}_{2}\\ {\xi}_{1}\end{array}\right).& & \end{array}$$  
Let
${\mathcal{T}}^{\prime}$
be the set of all the maximal trajectories
$(x,\sigma )$
whose switching signals
$\sigma $
are given by the feedback rule
$$\begin{array}{ccc}\sigma \left(t\right)=\{\begin{array}{ccc}1& \text{if}& {x}_{1}\left(t\right)<0\\ 2& \text{if}& {x}_{1}\left(t\right)\ge 0.\end{array}& & \end{array}$$  
Observe that the origin is a stable focus for the first subsystem and a center for the other, and that the trajectories of both are running counterclockwise.
Since the time needed by any nontrivial trajectory of the subsystem
$\dot{x}={f}_{1}\left(x\right)$
(
$\dot{x}={f}_{2}\left(x\right)$
) to go from the positive (resp. negative)
${x}_{2}$
axis to the negative (resp. positive) one is constant, clearly
${\mathcal{T}}^{\prime}\subseteq \mathcal{T}a[{\tau}_{D},1]$
for some
${\tau}_{D}>0$
. If we consider the closed covering of
${\mathbb{R}}^{2}:{\chi}_{1}=\{\xi :{\xi}_{1}\le 0\},{\chi}_{2}=\{\xi :{\xi}_{1}\ge 0\}$
, then
${\mathcal{T}}^{\prime}\subseteq \mathcal{T}\left[\chi \right]\cap \mathcal{T}a[{\tau}_{D},1]$
.
The function
$V:{\mathbb{R}}^{2}\times \{1,2\}\to \mathbb{R}$
defined by
$V(\xi ,i)=\xi {}^{2}$
is clearly a weak Lyapunovlike function for
${\mathcal{T}}^{\prime}$
. We claim that
$M=\left\{0\right\}\times \{1,2\}$
. In fact, let
$(\xi ,\gamma )\in M$
. Then there exists
$(x,\sigma )\in {\mathcal{T}}_{V}^{*}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a$
such that
$x\left(0\right)=\xi $
,
$\sigma \left(0\right)=\gamma $
and
$\left(x\right(t),\sigma (t\left)\right)\in M$
for all
$t\ge 0$
. From the facts that
$x\left(t\right)$
cannot remain forever in the right halfplane, where
$V\left(x\right(t),\sigma (t\left)\right)$
is constant, that
$V\left(x\right(t),\sigma (t\left)\right)$
is strictly decreasing when
$x\left(t\right)$
is in the open lefthalf plane, and from the definition of
${\mathcal{T}}_{V}^{*}$
, it follows readily that
$\left(x\right(t),\sigma (t\left)\right)$
cannot belong to
$M$
unless
$x\left(t\right)=0$
for all
$t\ge 0$
, and the claim follows. Hence, according to Theorem 2.2 ,
${\mathcal{T}}^{\prime}$
is globally uniformly asymptotically stable.
Example 2. Consider now the two systems in
${\mathbb{R}}^{2}$
given by:
$$\begin{array}{cc}\dot{x}={f}_{1}\left(x\right)=\left(\begin{array}{c}{x}_{1}{x}_{2}\\ {x}_{1}\end{array}\right)\text{and}\dot{x}={f}_{2}\left(x\right)=\frac{x}{1+x{}^{4}}.& \end{array}$$  
$$\begin{array}{}\end{array}$$  
Let
${W}_{1}\left(\xi \right)={\xi}_{1}^{2}$
,
${W}_{2}\left(\xi \right)=\frac{\xi {}^{2}}{1+\xi {}^{4}}$
and
$V(\xi ,i)=\xi {}^{2}/2,i=1,2$
. Then for every
$\xi \in {\mathbb{R}}^{2}$
,
$\frac{\partial V}{\partial \xi}(\xi ,1){f}_{1}\left(\xi \right)={W}_{1}\left(\xi \right)$
and
$\frac{\partial V}{\partial \xi}(\xi ,2){f}_{2}\left(\xi \right)={W}_{2}\left(\xi \right)$
. It is not hard to see that both pairs
$({f}_{1},{W}_{1})$
and
$({f}_{2},{W}_{2})$
have the zero smalltime distinguishability property and that the second one is not smalltime normobservable.
As a matter of fact it is not largetime norm observable. In this case Theorem 7 in [
6]
cannot be applied, but according to Theorem 2.3 , any
${\mathcal{T}}^{\prime}\subseteq \mathcal{T}a[{\tau}_{D},{N}_{0}]$
with fixed, but otherwise arbitrary,
${\tau}_{D}>0,{N}_{0}\in {\mathbb{N}}_{0}$
, is globally uniformly asymptotically stable.
It is worth noting that if one applied Theorem 1 of [
1]
, the attracting set so obtained would be the
${x}_{2}$
axis (see example 4 of [
1]
).
4 Invariance for switched systems
In this section we study the asymptotic behavior of bounded solutions
$x$
of the switching system ( 1 ) corresponding to switching signals which have a positive average dwelltime, and in particular the invariance properties of their
$\omega $
limit sets. We recall that a point
$\xi \in {\mathbb{R}}^{n}$
belongs to
$\Omega \left(x\right)$
, the
$\omega $
limit set of
$x:{\mathbb{R}}_{+}\to {\mathbb{R}}^{n}$
, if there exists a strictly increasing sequence of times
$\left\{{s}_{k}\right\}$
with
${lim}_{k\to \infty}{s}_{k}=\infty $
and
${lim}_{k\to \infty}x\left({s}_{k}\right)=\xi $
. The
$\omega $
limit set
$\Omega \left(x\right)$
is always closed and, when
$x$
is bounded, it is nonempty, compact and
${lim}_{t\to \infty}d\left(x\right(t),\Omega (x\left)\right)=0$
. Moreover,
$\Omega \left(x\right)$
is the smallest closed set which is approached by
$x$
.
In order to proceed, we will associate to each bounded trajectory
$(x,\sigma )\in \mathcal{T}a$
a nonempty subset of
${\mathbb{R}}^{n}\times \Gamma $
, which we denote
${\Omega}^{\u266f}(x,\sigma )$
, and study its invariance properties.
Let us introduce some more notation and terminology. As stated above, associated with a switching signal
$\sigma $
there are a strictly increasing sequence of real numbers (the sequence of switching times of
$\sigma $
)
$\{{t}_{i}{\}}_{i=0}^{{N}_{\sigma}}$
, with
${N}_{\sigma}$
finite or
${N}_{\sigma}=\infty $
,
${t}_{0}=0$
and
${lim}_{i\to \infty}{t}_{i}=\infty $
when
${N}_{\sigma}=\infty $
, and a sequence of points
$\{{\gamma}_{i}{\}}_{i=0}^{{N}_{\sigma}}\subseteq \Gamma $
, with
${\gamma}_{i}\ne {\gamma}_{i+1}$
for all
$0\le i<{N}_{\sigma}$
, such that
$\sigma \left(t\right)={\gamma}_{i}$
for all
${t}_{i}\le t<{t}_{i+1}$
with
$0\le i<{N}_{\sigma}$
, and
$\sigma \left(t\right)={\gamma}_{{N}_{\sigma}}$
for all
$t\ge {t}_{{N}_{\sigma}}$
when
${N}_{\sigma}$
is finite. In order to treat the cases
${N}_{\sigma}<\infty $
and
${N}_{\sigma}=\infty $
in an unified frame, we pick any
${\gamma}^{*}\in \Gamma $
and define
${t}_{i}=\infty $
and
${\gamma}_{i}={\gamma}^{*}$
for all
$i>{N}_{\sigma}$
when
${N}_{\sigma}$
is finite.
Given a switching signal
$\sigma \in \mathcal{S}$
we consider the sequence of maps
${\tau}^{i}\sigma :{\mathbb{R}}_{+}\cup \{\infty \}\to {\mathbb{R}}_{+}\cup \{\infty \}$
,
$i\in \mathbb{N}$
, defined recursively by:

∙
${\tau}^{1}\sigma \left(t\right)={t}_{k}$
if
$t\in [{t}_{k1},{t}_{k})$
and
${\tau}^{1}\sigma (\infty )=\infty $
;

∙
${\tau}^{i+1}\sigma \left(t\right)={\tau}^{1}\sigma \left({\tau}^{i}\sigma \right(t\left)\right)$
for all
$t\in {\mathbb{R}}_{+}\cup \{\infty \}$
and all
$i\ge 2$
.
We observe that for a given time
$t\ge 0$
,
${\tau}^{1}\sigma \left(t\right)$
is the first switching time greater than
$t$
,
${\tau}^{2}\sigma \left(t\right)$
is the second switching time greater than
$t$
, etc. We also define, for convenience,
${\tau}^{0}\sigma \left(t\right)=t$
for all
$t\ge 0$
.
Definition 4.1
Given a bounded trajectory
$(x,\sigma )\in \mathcal{T}a$
, a point
$(\xi ,\gamma )\in {\mathbb{R}}^{n}\times \Gamma $
belongs to
${\Omega}^{\u266f}(x,\sigma )$
if there exists a strictly increasing and unbounded sequence
$\left\{{s}_{k}\right\}\subset {\mathbb{R}}_{+}$
such that

1.
${lim}_{k\to \infty}{\tau}^{1}\sigma \left({s}_{k}\right){s}_{k}=r,0<r\le \infty $
;

2.
${lim}_{k\to \infty}x\left({s}_{k}\right)=\xi $
and
${lim}_{k\to \infty}\sigma \left({s}_{k}\right)=\gamma $
.
$\square $
We observe that in the case when
${N}_{\sigma}$
is finite,
${\Omega}^{\u266f}(x,\sigma )=\Omega \left(x\right)\times \left\{{\gamma}_{{N}_{\sigma}}\right\}$
.
The following lemma shows the relation between
$\Omega \left(x\right)$
and
${\Omega}^{\u266f}(x,\sigma )$
.
Lemma 4.1
Let
$(x,\sigma )$
be a bounded trajectory belonging to
$\mathcal{T}a$
. Then
$\Omega \left(x\right)={\pi}_{1}\left({\Omega}^{\u266f}\right(x,\sigma \left)\right)$
.
Proof. We only prove the case when
$\sigma $
has infinitely many switching times since the other case is trivial.
Suppose that
$\sigma \in \mathcal{S}a[{\tau}_{D},{N}_{0}]$
for some
${\tau}_{D}>0$
and
${N}_{0}\in \mathbb{N}$
and that
$\sigma $
has infinitely many switching times. We first note that the inclusion
${\pi}_{1}\left({\Omega}^{\u266f}\right(x,\sigma \left)\right)\subseteq \Omega \left(x\right)$
readily follows from the definition of
${\Omega}^{\u266f}(x,\sigma )$
.
In order to prove that
$\Omega \left(x\right)\subseteq {\pi}_{1}\left({\Omega}^{\u266f}\right(x,\sigma \left)\right)$
let
$\xi \in \Omega \left(x\right)$
. Then there exists a strictly increasing an unbounded sequence of times
$\left\{{s}_{k}\right\}$
such that
$x\left({s}_{k}\right)\to \xi $
.
Let
$i\ge 0$
be the first integer such that
$${limsup}_{k\to \infty}{\tau}^{i+1}\sigma \left({s}_{k}\right){s}_{k}=r,with0<r\le \infty .$$
Such an integer exists and verifies
$i\le {N}_{0}$
since, due to the definition of
$\mathcal{S}a[{\tau}_{D},{N}_{0}]$
,
${\tau}^{{N}_{0}+1}\sigma \left({s}_{k}\right){s}_{k}\ge {\tau}_{D}$
for all
$k\ge 1$
. Let
$\left\{{s}_{{k}_{j}}\right\}$
be a subsequence of
$\left\{{s}_{k}\right\}$
so that
${lim}_{j\to \infty}{\tau}_{\sigma}^{i+1}\left({s}_{{k}_{j}}\right){s}_{{k}_{j}}=r$
. Then, i)
${lim}_{j\to \infty}{\tau}^{l}\sigma \left({s}_{{k}_{j}}\right){s}_{{k}_{j}}=0$
for all
$0\le l\le i$
. Consider the sequence
$\left\{\sigma \right({\tau}^{i}\sigma \left({s}_{{k}_{j}}\right)){\}}_{j=1}^{\infty}$
; as
$\Gamma $
is compact, there is a subsequence
$\left\{\sigma \right({\tau}^{i}\sigma \left({s}_{{k}_{{j}_{l}}}\right)){\}}_{l=1}^{\infty}$
which converges to some
$\gamma \in \Gamma $
.
We claim that
$(\xi ,\gamma )\in {\Omega}^{\u266f}(x,\sigma )$
. In order to prove the claim, consider the unbounded sequence
$\{{s}_{l}^{\prime}={\tau}^{i}\sigma ({s}_{{k}_{{j}_{l}}}\left)\right\}$
. From i) and the facts that
${lim}_{j\to \infty}{\tau}^{i+1}\sigma \left({s}_{{k}_{j}}\right){s}_{{k}_{j}}=r$
and
${\tau}^{1}\sigma \left({s}_{l}^{\prime}\right)={\tau}^{1}\sigma \left({\tau}^{i}\sigma \right({s}_{{k}_{{j}_{l}}}\left)\right)={\tau}^{i+1}\sigma \left({s}_{{k}_{{j}_{l}}}\right)$
, we have that
${lim}_{l\to \infty}{\tau}^{1}\sigma \left({s}_{l}^{\prime}\right){s}_{l}^{\prime}=r>0$
.
We note that by construction
$${lim}_{l\to \infty}\sigma \left({s}_{l}^{\prime}\right)={lim}_{l\to \infty}\sigma \left({\tau}^{i}\sigma \right({s}_{{k}_{{j}_{l}}}\left)\right)=\gamma .$$
Finally, taking into account that:

∙
${lim}_{l\to \infty}{s}_{l}^{\prime}{s}_{{k}_{{j}_{l}}}=0$
;

∙
$x$
is uniformly continuous on
${\mathbb{R}}_{+}$
since
$\dot{x}$
is essentially bounded (
$x$
is bounded,
$f$
is continuous and
$\Gamma $
is compact);

∙
${lim}_{k\to \infty}x\left({s}_{k}\right)=\xi $
;
it follows that
${lim}_{l\to \infty}x\left({s}_{l}^{\prime}\right)=\xi $
. Thus
$(\xi ,\gamma )\in {\Omega}^{\u266f}(x,\sigma )$
and thereby
$\xi \in {\pi}_{1}\left({\Omega}^{\u266f}\right(x,\sigma \left)\right)$
. Then
$\Omega \left(x\right)\subseteq {\pi}_{1}\left({\Omega}^{\u266f}\right(x,\sigma \left)\right)$
and the lemma follows.
The next result shows that, under suitable hypotheses, the set
${\Omega}^{\u266f}(x,\sigma )$
corresponding to a trajectory in
${\mathcal{T}}_{V}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a[{\tau}_{D},{N}_{0}]$
is weaklyinvariant w.r.t.
${\mathcal{T}}_{V}^{*}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a[{\tau}_{D},{N}_{0}]$
.
Proposition 4.1
Let
$\chi $
be a closed covering of
${\mathbb{R}}^{n}$
and let
$V:\Omega \times \Gamma \to \mathbb{R}$
be a continuous function. Suppose that
$(x,\sigma )$
is a trajectory belonging to
${\mathcal{T}}_{V}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a[{\tau}_{D},{N}_{0}]$
,
${\tau}_{D}>0,{N}_{0}\in \mathbb{N}$
, such that for some compact set
$B\subset \Omega $
,
$x\left(t\right)\in B$
for all
$t\ge 0$
. Then
${\Omega}^{\u266f}(x,\sigma )$
is weaklyinvariant w.r.t.
${\mathcal{T}}_{V}^{*}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a[{\tau}_{D},{N}_{0}]$
.
$\square $
Proof. Let
$(\xi ,\gamma )\in {\Omega}^{\u266f}(x,\sigma )$
. Then there exists a strictly increasing and unbounded sequence
$\left\{{s}_{k}\right\}$
which verifies 1. and 2. of Definition 4.1 . Let
${\sigma}_{k}(\cdot )=\sigma (\cdot +{s}_{k})$
and
${x}_{k}(\cdot )=x(\cdot +{s}_{k})$
. As
$\chi $
,
$V$
and the sequence
$\left\{\right({x}_{k},{\sigma}_{k}\left)\right\}$
are as in the hypotheses of Lemma A.2 in the Appendix, there exist a subsequence
$({x}_{{k}_{l}},{\sigma}_{{k}_{l}})$
and a trajectory
$({x}^{*},{\sigma}^{*})\in {\mathcal{T}}_{V}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a[{\tau}_{D},{N}_{0}]$
such that
$\left\{{x}_{{k}_{l}}\right\}$
converges uniformly to
${x}^{*}$
on compact subsets of
${\mathbb{R}}_{+}$
and
$\left\{{\sigma}_{{k}_{l}}\right\}$
converges to
${\sigma}^{*}$
a.e. on
${\mathbb{R}}^{+}$
.
Due to Lemma
A.1 in the Appendix, we can also assume without loss of generality that
$\left\{{\sigma}_{{k}_{l}}\right\}$
also verifies condition 2 of that lemma with
${\sigma}^{*}$
in place of
$\sigma $
.
The proof is completed provided we show that
$\left({x}^{*}\right(0),{\sigma}^{*}(0\left)\right)=(\xi ,\gamma )$
,
$\left({x}^{*}\right(t),{\sigma}^{*}(t\left)\right)\in {\Omega}^{\u266f}(x,\sigma )$
for all
$t\ge 0$
and
$({x}^{*},{\sigma}^{*})\in {\mathcal{T}}_{V}^{*}$
.
Let us prove first that
$\left({x}^{*}\right(0),{\sigma}^{*}(0\left)\right)=(\xi ,\gamma )$
. From the fact that
${x}_{k}\left(0\right)=x\left({s}_{k}\right)$
, 2. of Definition 4.1 and the convergence of
$\left\{{x}_{{k}_{l}}\right\}$
to
${x}^{*}$
, we have that
${x}^{*}\left(0\right)=\xi $
.
According to 2. of Lemma
A.1 , there exists a sequence
$\left\{{r}_{l}\right\}\subset {\mathbb{R}}_{+}$
such that
${lim}_{l\to \infty}{r}_{l}=0$
,
${lim}_{l\to \infty}{\tau}_{{\sigma}_{{k}_{l}}}^{1}\left({r}_{l}\right){r}_{l}>0$
and
${lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left({r}_{l}\right)={\sigma}^{*}\left(0\right)$
. From item 1. of Definition 4.1 and the fact that
${\tau}_{{\sigma}_{k}}^{1}\left(0\right)={\tau}_{\sigma}^{1}\left({s}_{k}\right){s}_{k}$
for all
$k\in \mathbb{N}$
, it follows that
${lim}_{l\to \infty}{\tau}_{{\sigma}_{{k}_{l}}}^{1}\left(0\right)>0$
. Then
${r}_{l}<{\tau}_{{\sigma}_{{k}_{l}}}^{1}\left(0\right)$
for
$l$
large enough and, therefore,
${\sigma}^{*}\left(0\right)={lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left({r}_{l}\right)={lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left(0\right)={lim}_{l\to \infty}\sigma \left({s}_{{k}_{l}}\right)=\gamma $
.
Next we prove that
$\left({x}^{*}\right(t),{\sigma}^{*}(t\left)\right)\in {\Omega}^{\u266f}(x,\sigma )$
for all
$t>0$
.
Let
$t>0$
and let
$\left\{{r}_{l}\right\}$
be a sequence as in 2. of Lemma A.1 . Consider the unbounded sequence
$\left\{{s}_{l}^{\prime}\right\}$
, defined by
${s}_{l}^{\prime}={r}_{l}+{s}_{{k}_{l}}$
, which we can suppose, without loss of generality, strictly increasing. Due to the fact that
${\tau}_{\sigma}^{1}\left({s}_{l}^{\prime}\right)={\tau}_{{\sigma}_{{k}_{l}}}^{1}\left({r}_{l}\right)+{s}_{{k}_{l}}$
, we have that
${lim}_{l\to \infty}{\tau}_{\sigma}^{1}\left({s}_{l}^{\prime}\right){s}_{l}^{\prime}>0$
. So
$\left\{{s}_{l}^{\prime}\right\}$
satisfies condition 1. of Definition 4.1 . From 2. of Lemma A.1 , we have that
${lim}_{l\to \infty}\sigma \left({s}_{l}^{\prime}\right)={lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left({r}_{l}\right)={\sigma}^{*}\left(t\right)$
. On the other hand, from the uniform convergence of
$\left\{{x}_{{k}_{l}}\right\}$
to
${x}^{*}$
on compact sets and the continuity of
${x}^{*}$
we have that
${lim}_{l\to \infty}x\left({s}_{l}^{\prime}\right)={lim}_{l\to \infty}{x}_{{k}_{l}}\left({r}_{l}\right)={x}^{*}\left(t\right)$
. Hence
$\left(x\right({s}_{l}^{\prime}),\sigma ({s}_{l}^{\prime}\left)\right)\to \left({x}^{*}\right(t),{\sigma}^{*}(t\left)\right)$
as
$l\to \infty $
and thereby
$\left({x}^{*}\right(t),{\sigma}^{*}(t\left)\right)\in {\Omega}^{\u266f}(x,\sigma )$
.
Finally, we prove that
$({x}^{*},{\sigma}^{*})\in {\mathcal{T}}_{V}^{*}$
. Since
$({x}^{*},{\sigma}^{*})\in {\mathcal{T}}_{V}$
it suffices to prove that for any pair of times
$t,{t}^{\prime}$
with
$t<{t}^{\prime}$
and
${\sigma}^{*}\left(t\right)={\sigma}^{*}\left({t}^{\prime}\right)$
,
$V\left({x}^{*}\right(t),{\sigma}^{*}(t\left)\right)\le V\left({x}^{*}\right({t}^{\prime}),{\sigma}^{*}({t}^{\prime}\left)\right)$
.
Let
$t,{t}^{\prime}$
be a pair of times such that
$t<{t}^{\prime}$
and
${\sigma}^{*}\left(t\right)={\sigma}^{*}\left({t}^{\prime}\right)=\gamma $
. Since
${lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left(t\right)={\sigma}^{*}\left(t\right)$
a.e. on
${\mathbb{R}}_{+}$
, and
${\sigma}^{*}$
is piecewise constant and right continuous, there exists a pair of nonincreasing sequences
$\left\{{\tau}_{i}\right\}$
,
$\left\{{\tau}_{i}^{\prime}\right\}$
so that

∙
${\tau}_{i}<{\tau}_{i}^{\prime}$
for all
$i$
;

∙
${\tau}_{i}\searrow t$
and
${\tau}_{i}^{\prime}\searrow {t}^{\prime}$
;

∙
for every
$i$
,
${lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left({\tau}_{i}\right)={\sigma}^{*}\left({\tau}_{i}\right)=\gamma $
and
${lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left({\tau}_{i}^{\prime}\right)={\sigma}^{*}\left({\tau}_{i}^{\prime}\right)=\gamma $
.
Fix
$i$
. Since
$\Gamma $
is finite, there exists
${l}^{*}$
such that
${\sigma}_{{k}_{l}}\left({\tau}_{i}\right)={\sigma}_{{k}_{l}}\left({\tau}_{i}^{\prime}\right)=\gamma $
for all
$l\ge {l}^{*}$
. Fix
${l}^{\prime}\ge {l}^{*}$
. As
$\left\{{s}_{{k}_{l}}\right\}$
is unbounded, we have that
${\tau}_{i}+{s}_{{k}_{l}}>{\tau}_{i}^{\prime}+{s}_{{k}_{{l}^{\prime}}}$
for
$l$
large enough, say
$l\ge {l}_{0}$
. Then, for
$l\ge max\{{l}_{0},{l}^{*}\}$
,
${\tau}_{i}+{s}_{{k}_{l}}>{\tau}_{i}^{\prime}+{s}_{{k}_{{l}^{\prime}}}$
and
$\sigma ({\tau}_{i}+{s}_{{k}_{l}})={\sigma}_{{k}_{l}}\left({\tau}_{i}\right)={\sigma}_{{k}_{{l}^{\prime}}}\left({\tau}_{i}^{\prime}\right)=\sigma ({\tau}_{i}^{\prime}+{s}_{{k}_{{l}^{\prime}}})$
. In consequence,
$$\begin{array}{ccc}V\left({x}_{{k}_{{l}^{\prime}}}\right({\tau}_{i}^{\prime}),{\sigma}_{{k}_{{l}^{\prime}}}({\tau}_{i}^{\prime}\left)\right)=V\left(x\right({\tau}_{i}^{\prime}+{s}_{{k}_{{l}^{\prime}}}),\sigma ({\tau}_{i}^{\prime}+{s}_{{k}_{{l}^{\prime}}}\left)\right)\ge & & \end{array}$$  
$$\begin{array}{ccc}V\left(x\right({\tau}_{i}+{s}_{{k}_{l}}),\sigma ({\tau}_{i}+{s}_{{k}_{l}}\left)\right)=V\left({x}_{{k}_{l}}\right({\tau}_{i}),{\sigma}_{{k}_{l}}({\tau}_{i}\left)\right).& & \end{array}$$  
From the latter, after taking limit as
$l\to \infty $
, we get
$V\left({x}_{{k}_{{l}^{\prime}}}\right({\tau}_{i}^{\prime}),{\sigma}_{{k}_{{l}^{\prime}}}({\tau}_{i}^{\prime}\left)\right)\ge V\left({x}^{*}\right({\tau}_{i}),{\sigma}^{*}({\tau}_{i}\left)\right)$
and from this, letting
${l}^{\prime}\to \infty $
, we obtain
$V\left({x}^{*}\right({\tau}_{i}^{\prime}),{\sigma}^{*}({\tau}_{i}^{\prime}\left)\right)\ge V\left({x}^{*}\right({\tau}_{i}),{\sigma}^{*}({\tau}_{i}\left)\right)$
. Finally, from the continuity of
$V$
and
${x}^{*}$
and the right continuity of
${\sigma}^{*}$
, letting
$i\to \infty $
it follows that
$V\left({x}^{*}\right(t),{\sigma}^{*}(t\left)\right)\le V\left({x}^{*}\right({t}^{\prime}),{\sigma}^{*}({t}^{\prime}\left)\right)$
.
Now we are ready to prove the extension of LaSalle's invariance principle given in Theorem
2.4 .
Proof of Theorem 2.4 . It readily follows from Proposition 4.1 and Lemma 4.1 . In fact, from Proposition 4.1 we have that
${\Omega}^{\u266f}(x,\sigma )\subseteq M$
. Thus, from Lemma 4.1 , we deduce that
$\Omega \left(x\right)\subseteq {\pi}_{1}\left(M\right)$
and therefore that
$x\left(t\right)$
tends to
${\pi}_{1}\left(M\right)$
as
$t\to \infty $
.
5 Proof of Theorem 2.2
Proof of Theorem 2.2 . Since from Theorem 2.1 we know that
${\mathcal{T}}^{\prime}$
is uniform stable, we only have to prove the remaining statements.
Let
$(x,\sigma )\in {\mathcal{T}}^{\prime}$
. Since
${\mathcal{T}}^{\prime}$
is uniform stable, it follows that
$(x,\sigma )$
is bounded and therefore evolves into some compact subset
$B$
of
$\Omega ={\mathbb{R}}^{n}$
. By applying Theorem 2.4 we deduce that
$x\left(t\right)$
tends to
${\pi}_{1}\left(M\right)=\left\{0\right\}$
and the global asymptotic stability of
${\mathcal{T}}^{\prime}$
follows.
Suppose now that
${\mathcal{T}}^{\prime}\subseteq \mathcal{T}a[{\tau}_{D},{N}_{0}]$
. Since
$V$
is a weak Lyapunovlike function for the family of trajectories
${\mathcal{T}}^{*}={\mathcal{T}}_{V}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a[{\tau}_{D},{N}_{0}]$
, and
${\mathcal{T}}^{\prime}\subseteq {\mathcal{T}}^{*}$
, it suffices to prove that
${\mathcal{T}}^{*}$
is globally uniformly asymptotically stable.
Since we have already proved that
${\mathcal{T}}^{*}$
is globally asymptotically stable, and in particular uniformly stable, the global uniform asymptotic stability of
${\mathcal{T}}^{*}$
will be established if we show that
${\mathcal{T}}^{*}$
verifies (a) of Remark 2.1 .
As
${\mathcal{T}}^{*}$
is invariant by time translations, i.e., for all
$s\ge 0$
,
$\left(x\right(\cdot +s),\sigma (\cdot +s\left)\right)\in {\mathcal{T}}^{*}$
if
$(x,\sigma )\in {\mathcal{T}}^{*}$
, in order to prove (a) of Remark 2.1 it is sufficient to show that
${\mathcal{T}}^{*}$
verifies the weaker condition:

(*)
for each
$R>0$
and each
$\varepsilon >0$
, there exists
$T\ge 0$
such that for all
$(x,\sigma )\in {\mathcal{T}}^{*}$
,
$$\leftx\right(0\left)\right<R\u27f9\leftx\right(t\left)\right<\varepsilon \forall t\ge T.$$
Suppose that (*) does not hold. Then there exist
${\varepsilon}_{0}>0$
,
${\eta}_{0}>0$
, a sequence of trajectories
$\left\{\right({x}_{k},{\sigma}_{k}\left)\right\}\subset {\mathcal{T}}^{*}$
and an increasing and unbounded sequence of times
$\left\{{\tau}_{k}\right\}$
such that
$\left{x}_{k}\right(0\left)\right\le {\eta}_{0}$
and
$\left{x}_{k}\right({\tau}_{k}\left)\right\ge {\varepsilon}_{0}$
for all
$k$
.
Since
$\left\{\right({x}_{k},{\sigma}_{k}\left)\right\}$
is uniformly bounded, from Lemma A.2 we know that there exists a subsequence
$\left\{\right({x}_{{k}_{l}},{\sigma}_{{k}_{l}}\left)\right\}$
and a trajectory
$({x}^{*},{\sigma}^{*})\in {\mathcal{T}}^{*}$
such that
$\left\{{x}_{{k}_{l}}\right\}$
converges to
${x}^{*}$
uniformly on compact sets.
Let
${\varepsilon}^{\prime}={\alpha}^{1}({\varepsilon}_{0}/2)$
, with
$\alpha $
as in 1. of Definition 2.1 . As
${x}^{*}$
converges to
$0$
as
$t\to \infty $
, there exists a time
$T>0$
such that
$\left{x}^{*}\right(T\left)\right<{\varepsilon}^{\prime}$
. Since
$\left\{{x}_{{k}_{l}}\right\}$
converges to
${x}^{*}$
uniformly on compact sets,
$\left{x}_{{k}_{l}}\right(T\left)\right<{\varepsilon}^{\prime}$
for
$l$
large enough. Then, due to the uniform stability of
${\mathcal{T}}^{*}$
, we have that, for
$l$
large enough and
$t\ge T$
,
$$\left{x}_{{k}_{l}}\right(t\left)\right\le \alpha \left(\right{x}_{{k}_{l}}\left(T\right)\left\right)\le \alpha \left({\varepsilon}^{\prime}\right)=\frac{{\varepsilon}_{0}}{2},$$
which is a contradiction.
6 Conclusions
In this paper we have presented an extension of LaSalle's invariance principle for switched nonlinear systems assuming that the family of subsystems is finite and that the switching signals have a positive average dwelltime. This extension enabled us to obtain some asymptotic stability criteria for this class of systems. Examples were presented that show the application of our results to cases either intractable with some of the previously mentioned results or upon which those results give no conclusive answers.
In addition, results about the compactness of the trajectories of the systems involved were exhibit that not only were instrumental used in the proof of some results along the paper, but are important by their own.
Finally, we point out that extensions of LaSalle's principle for switched nonlinear systems in the case that
$\Gamma $
is infinite and also integral invariance principles for the same class of switched systems have been already obtained and are currently under preparation for publication.
References

A. Bacciotti and L. Mazzi, An invariance principle for nonlinear switched systems, Rapporti interni Nr. 22, Departimento di Matematica, Politecnico di Torino, 2004.

M. Branicky, Multiple Lyapunov functions and other analysis tools for switched and hybrid systems, IEEE Trans. Automat. Contr., vol. 43, 475482, 1998.

R.A. DeCarlo, M.S. Branicky, S. Pettersson and B. Lennartson, Perspectives and results on the stability and stabilizability of hybrid systems, Proc. IEEE, vol. 88, pp. 10691082, 2000.

J. Dugundji, Topology, Allyn and Bacon, Inc., Boston, 1966.

J.P Hespanha, Uniform stability properties of switched linear systems: extensions of LaSalle's Invariance Principle, IEEE Trans. Automat. Control, vol. 49, pp. 470482, 2004.

J.P. Hespanha, D. Liberzon, D. Angeli and E. Sontag, Nonlinear observability notions and stability of switched systems, IEEE Trans. Automat. Control, vol. 50, pp. 154168, 2005.

P.E. Kloeden, Nonautonomous attractors of switched systems, submitted, 2004.

J. P. LaSalle, The Stability of Dynamical Systems, ser. Regional Conference Series in Applied Mathematics, Philadelphia, PA: SIAM, 1996.

D. Liberzon, Switching in Systems and Control, Birkhäuser, Boston, 2003.

D. Liberzon and A.S. Morse, Basic problems in stability and design of switched systems, IEEE Control Systems Magazine, vol. 19, pp. 5970, 1999.

Y. Lin, E.D. Sontag and Y. Wang, A smooth converse Lyapunov theorem for robust stability, SIAM J. Control Optim., vol. 34, pp. 124160, 1996.

J. Lygeros, K. H. Johansson, S. M. Simić, J. Zhang and S. Sastry, Dynamical Properties of Hybrid Automata, IEEE Trans. Automat. Contr. vol. 48, pp. 217, 2003.

P. Peleties and R.A. DeCarlo, Asymptotic stability of
$m$
switched systems using Lyapunovlike functions, in Proc. 1991 Amer. Control Conf., 1991. pp. 16791684.
Appendix
A Some compactness results of tra jectories of switched systems
In this Appendix we will show that, under suitable hypotheses, certain families of trajectories of system ( 1 ) enjoy a certain kind of sequential compactness.
We say that a sequence
$\left\{\right({x}_{k},{\sigma}_{k}\left)\right\}$
of trajectories of ( 1 ) is uniformly bounded if there exists
$M\ge 0$
such that for all
$k$
,
$\left{x}_{k}\right(t\left)\right\le M$
for all
$t\ge 0$
.
Definition A.1
A family
${\mathcal{T}}^{*}$
of maximal trajectories of ( 1 ) has the SC property if for every uniformly bounded sequence
$\left\{\right({x}_{k},{\sigma}_{k}\left)\right\}\subset {\mathcal{T}}^{*}$
there exist a subsequence
$\left\{\right({x}_{{k}_{l}},{\sigma}_{{k}_{l}}\left)\right\}$
and a trajectory
$({x}^{*},{\sigma}^{*})\in {\mathcal{T}}^{*}$
such that
$\left\{{x}_{{k}_{l}}\right\}$
converges to
${x}^{*}$
uniformly on compact sets of
${\mathbb{R}}_{+}$
and
${lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left(t\right)={\sigma}^{*}\left(t\right)$
a.e. on
${\mathbb{R}}_{+}$
.
$\square $
Proposition A.1
Assume that
$\chi $
is a closed covering of
${\mathbb{R}}^{n}$
.Then
$\mathcal{T}\left[\chi \right]\cap \mathcal{T}a[{\tau}_{D},{N}_{0}]$
has the SC property for all
${\tau}_{D}>0$
and all
${N}_{0}\in \mathbb{N}$
.
$\square $
The following lemma is used in the proof of Proposition A.1 and in some parts of section 4.
Lemma A.1
Let
$\left\{{\sigma}_{k}\right\}$
be a sequence of switching signals in
$\mathcal{S}a[{\tau}_{D},{N}_{0}]$
with
${\tau}_{D}>0$
and
${N}_{0}\in \mathbb{N}$
.
Then there exist a subsequence
$\left\{{\sigma}_{{k}_{l}}\right\}$
and a switching signal
$\sigma \in \mathcal{S}a[{\tau}_{D},{N}_{0}]$
such that

1.
${lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left(t\right)=\sigma \left(t\right)$
for almost all
$t\ge 0$
;

2.
for each
$t\in {\mathbb{R}}_{+}$
there exists a sequence of positive times
$\{{r}_{l}{\}}_{l=1}^{\infty}$
such that
${lim}_{l\to \infty}{r}_{l}=t$
,
$${lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left({r}_{l}\right)=\sigma \left(t\right)and{lim}_{l\to \infty}{\tau}_{{\sigma}_{{k}_{l}}}^{1}\left({r}_{l}\right){r}_{l}>0.$$
$\square $
Proof. Let
$\overline{\mathbb{R}+}={\mathbb{R}}_{+}\cup \{\infty \}$
be the onepoint compactification of
${\mathbb{R}}_{+}$
, which we recall is a compact metric space, and let
$\mathcal{K}=(\overline{\mathbb{R}+}\times \Gamma {)}^{{\mathbb{N}}_{0}}$
be the set of all the sequences
$p=\left\{\right({t}_{i},{\gamma}_{i}):{t}_{i}\in \overline{\mathbb{R}+},{\gamma}_{i}\in \Gamma ,i\in {\mathbb{N}}_{0}\}$
endowed with the product topology. We note that since
$\mathcal{K}$
is the Cartesian product of a countable number of compact metric spaces, it is metrizable (see [4] , Theorem 7.2 on p. 190) and compact (see [4] , Theorem 1.4 on p. 224 ).
For each
$k\in \mathbb{N}$
, let
$\{{t}_{i}^{k}{\}}_{i=0}^{\infty}$
be the sequence of switching times associated to the switching signal
${\sigma}_{k}$
and let
$\{{\gamma}_{i}^{k}{\}}_{i=0}^{\infty}$
be the sequence of points of
$\Gamma $
defined by
${\gamma}_{i}^{k}={\sigma}_{k}\left({t}_{i}^{k}\right)$
. Observe that when
${N}_{{\sigma}_{k}}$
is finite we have, according to the convention above, that
${\gamma}_{i}^{k}={\gamma}^{*}$
and
${t}_{i}^{k}=\infty $
for every
$i>{N}_{{\sigma}_{k}}$
. Let
${p}_{k}=\left\{{p}_{k}\right(i)=({t}_{i}^{k},{\gamma}_{i}^{k}){\}}_{i=0}^{\infty}\in \mathcal{K}$
.
As
$\mathcal{K}$
is a compact metric space, there exists a subsequence
$\{{p}_{{k}_{l}}{\}}_{l=1}^{\infty}$
which converges, say to
$p=\left\{p\right(i)=({t}_{i},{\gamma}_{i}){\}}_{i=0}^{\infty}$
, i.e., for each
$i\in {\mathbb{N}}_{0}$
,
${t}_{i}^{{k}_{l}}\to {t}_{i}$
and
${\gamma}_{i}^{{k}_{l}}\to {\gamma}_{i}$
as
$l\to \infty $
.
Since for each
$k\in \mathbb{N}$
,
$\{{t}_{i}^{k}{\}}_{i=0}^{\infty}$
is nondecreasing and
${t}_{0}^{k}=0$
, it readily follows that
$\{{t}_{i}{\}}_{i=0}^{\infty}$
is nondecreasing and
${t}_{0}=0$
.
We claim that:

(a)
For every open interval
$(a,b)$
, with
$a<b$
, the number of indexes
$i\in \mathbb{N}$
such that
${t}_{i}\in (a,b)$
, is bounded by
${N}_{0}+(ba)/{\tau}_{D}$
;

(b)
the number of indexes
$i$
such that
${t}_{i}=0$
is at most
${N}_{0}+1$
.
We will only prove (a) since (b) can be proved in a similar way.
Proof of (a). Suppose on the contrary that there are
$r>{N}_{0}+(ba)/{\tau}_{D}$
indexes, say
${i}_{1},\dots ,{i}_{r}$
, such that
${t}_{{i}_{j}}\in (a,b)$
for
$j=1,\dots ,r$
. Since
${t}_{{i}_{j}}^{{k}_{l}}\to {t}_{{i}_{j}}$
for each
$j=1,\dots ,r$
, we have that
${t}_{{i}_{j}}^{{k}_{l}}\in (a,b)$
for all
$j=1,\dots ,r$
if
$l$
is large enough, which contradicts the fact that
${\sigma}_{k}$
belongs to
$\mathcal{S}a[{\tau}_{D},{N}_{0}]$
.
In order to define
$\sigma $
as in the thesis of the lemma, let
$\{{i}_{j}{\}}_{j=0}^{N}$
, with
$N\le \infty $
, be the unique subsequence of
${\mathbb{N}}_{0}$
that verifies:

∙
$0={t}_{0}=\cdots ={t}_{{i}_{0}}$
and
${t}_{{i}_{0}+1}>0$
;

∙
${t}_{{i}_{j}+1}=\cdots ={t}_{{i}_{j+1}}$
and
${t}_{{i}_{j+1}}<{t}_{{i}_{j+1}+1}$
for all
$0\le j<N$
;

∙
${t}_{{i}_{N}+1}=\infty $
when
$N<\infty $
.
Note that this subsequence is well defined due to (a) and (b).
Now, let
$\sigma :{\mathbb{R}}_{+}\to \Gamma $
be the switching signal defined by:
$\sigma \left(t\right)={\gamma}_{{i}_{j}}$
for all
$t\in [{t}_{{i}_{j}},{t}_{{i}_{j+1}})$
and all
$j\le N$
. From (a) it readily follows that
$\sigma \in \mathcal{S}a[{\tau}_{D},{N}_{0}]$
.
Now we proceed to prove 1. and 2. of the thesis of the lemma. We will consider two cases.
Case I.
$t/\in \{{t}_{{i}_{j}}{\}}_{j=0}^{N}$
.
Let
${j}^{*}\in {\mathbb{N}}_{0}$
so that
${t}_{{i}_{{j}^{*}}}<t<{t}_{{i}_{{j}^{*}}+1}$
. As
${lim}_{l\to \infty}{t}_{{i}_{{j}^{*}}}^{{k}_{l}}={t}_{{i}_{{j}^{*}}}$
and
${lim}_{l\to \infty}{t}_{{i}_{{j}^{*}}+1}^{{k}_{l}}={t}_{{i}_{{j}^{*}}+1}$
,
${t}_{{i}_{{j}^{*}}}^{{k}_{l}}<t<{t}_{{i}_{{j}^{*}}+1}^{{k}_{l}}$
for
$l$
large enough, say
$l\ge L$
. Therefore, for
$l\ge L$
,
${\sigma}_{{k}_{l}}\left(t\right)={\sigma}_{{k}_{l}}\left({t}_{{i}_{{j}^{*}}}^{{k}_{l}}\right)={\gamma}_{{i}_{{j}^{*}}}^{{k}_{l}}$
and, consequently,
$$\begin{array}{ccc}{lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left(t\right)={lim}_{l\to \infty}{\gamma}_{{i}_{{j}^{*}}}^{{k}_{l}}={\gamma}_{{i}_{{j}^{*}}}=\sigma \left(t\right).& & \end{array}$$ 
(3)

As
$\{{t}_{{i}_{j}}{\}}_{j=0}^{N}$
is a set of measure zero, the latter shows 1).
From the fact that
${t}_{{i}_{{j}^{*}}}^{{k}_{l}}<t<{t}_{{i}_{{j}^{*}}+1}^{{k}_{l}}$
for
$l\ge L$
, we also have that
$$\begin{array}{ccc}{lim}_{l\to \infty}{\tau}_{{\sigma}_{{k}_{l}}}^{1}\left(t\right)t={lim}_{l\to \infty}{t}_{{i}_{{j}^{*}}+1}^{{k}_{l}}t={t}_{{i}_{{j}^{*}}+1}t>0,& & \end{array}$$  
which shows that 2) holds with the sequence
$\left\{{r}_{l}\right\}$
defined by
${r}_{l}=t$
for all
$l\in \mathbb{N}$
.
Case II.
$t={t}_{{i}_{{j}^{*}}}$
for some
${j}^{*}\in {\mathbb{N}}_{0}$
.
By using arguments similar to those used in the preceding case, we have that
$${lim}_{l\to \infty}{\tau}_{{\sigma}_{{k}_{l}}}^{1}\left(t\right){t}_{{i}_{{j}^{*}}}^{{k}_{l}}={lim}_{l\to \infty}{t}_{{i}_{{j}^{*}}+1}^{{k}_{l}}{t}_{{i}_{{j}^{*}}}^{{k}_{l}}={t}_{{i}_{{j}^{*}}+1}{t}_{{i}_{{j}^{*}}}>0,$$
and that
${lim}_{l\to \infty}\sigma \left({t}_{{i}_{{j}^{*}}}^{{k}_{l}}\right)=\sigma \left(t\right)$
.
Consequently, item 2. holds with the sequence
$\left\{{r}_{l}\right\}$
defined by
${r}_{l}={t}_{{i}_{{j}^{*}}}^{{k}_{l}}$
for all
$l\in \mathbb{N}$
.
Remark A.1
It is worth mentioning that it is not necessary that
$\Gamma $
be a finite set for the thesis of Lemma A.1 to hold. In fact, as can be easily seen from its proof, it suffices that
$\Gamma $
be a compact metric space.
$\square $
Remark A.2
A result on compactness of switching signals, proved with arguments different to ours, has recently appeared in [
7]
. That result (Theorem 1 of that paper) states that given
$\Gamma =\{1,...,N\}$
, the set of switching signals with a fixed dwelltime
${\tau}_{D}>0$
is a compact subset of the metric space
$(\mathcal{S},d)$
where the metric
$d$
is defined by
$$\begin{array}{ccc}d(u,v)={\sum}_{n=1}^{\infty}{2}^{n}{\int}_{0}^{n}\leftu\right(s)v(s\left)\rightds,& & \end{array}$$  
for all
$u$
and
$v$
in
$\mathcal{S}$
.
By using Lemma
A.1 one can generalize Theorem 1 in [
7]
to average dwelltime signals (the fixed dwelltime hypothesis is essential in the proof given in [
7]
). In fact, if we consider the metric space
$(\mathcal{S}a[{\tau}_{D},{N}_{0}],d)$
with
$d$
the metric above, the compactness of
$\mathcal{S}a[{\tau}_{D},{N}_{0}]$
follows from Lemma A.1 and the application of Lebesgue's Dominated Convergence Theorem. Moreover, since as pointed out in Remark A.1 , Lemma A.1 holds for a compact metric space
$(\Gamma ,\rho )$
,
$\mathcal{S}a[{\tau}_{D},{N}_{0}]$
is compact with the metric
$$\begin{array}{ccc}d(u,v)={\sum}_{n=1}^{\infty}{2}^{n}{\int}_{0}^{n}\rho \left(u\right(s),v(s\left)\right)ds,& & \end{array}$$  
for all
$u$
and
$v$
in
$\mathcal{S}$
.
$\square $
Proof of Proposition A.1 . As
$\left\{\right({x}_{k},{\sigma}_{k}\left)\right\}$
is bounded there exists
$M\ge 0$
such that
$\left{x}_{k}\right(t\left)\right\le M$
for all
$t\ge 0$
and all
$k\in \mathbb{N}$
. Let
${M}^{\prime}={max}_{\left\xi \right\le M,\gamma \in \Gamma}\leftf\right(\xi ,\gamma \left)\right$
. Then, for every positive integer
$k$
,
$\left{\dot{x}}_{k}\right(t\left)\right\le {M}^{\prime}$
for almost all
$t\in [0,+\infty )$
. In consequence
$\left\{{x}_{k}\right\}$
is equibounded and equicontinuous. Then, applying the ArzelaAscoli Theorem we deduce the existence of a subsequence
$\left\{{x}_{{k}_{l}}\right\}$
and a continuous function
${x}^{*}:{\mathbb{R}}_{+}\to {\mathbb{R}}^{n}$
such that
$\left\{{x}_{{k}_{l}}\right\}$
converges to
${x}^{*}$
uniformly on compact subsets of
${\mathbb{R}}_{+}$
.
Consider the subsequence
$\left\{{\sigma}_{{k}_{l}}\right\}$
. Due to Lemma A.1 we can suppose without loss of generality that there exists
${\sigma}^{*}\in \mathcal{S}a[{\tau}_{D},{N}_{0}]$
such that
${lim}_{l\to +\infty}{\sigma}_{{k}_{l}}\left(t\right)={\sigma}^{*}\left(t\right)$
for almost all
$t\ge 0$
.
We claim that
${x}^{*}$
is a solution of ( 1 ) corresponding to
${\sigma}^{*}$
and, in consequence,
$({x}^{*},{\sigma}^{*})\in \mathcal{T}a[{\tau}_{D},{N}_{0}]$
.
Let
$t>0$
. Then
$$\begin{array}{ccc}{x}^{*}\left(t\right)={lim}_{l\to +\infty}{x}_{{k}_{l}}\left(t\right)& =& {lim}_{l\to +\infty}\left({x}_{{k}_{l}}\left(0\right)+{\int}_{0}^{t}f\left({x}_{{k}_{l}}\right(s),{\sigma}_{{k}_{l}}(s\left)\right)ds\right)\end{array}$$  
$$\begin{array}{ccc}& =& {x}^{*}\left(0\right)+{lim}_{l\to +\infty}{\int}_{0}^{t}f\left({x}_{{k}_{l}}\right(s),{\sigma}_{{k}_{l}}(s\left)\right)ds.\end{array}$$  
As
$\leftf\right({x}_{{k}_{l}}\left(s\right),{\sigma}_{{k}_{l}}\left(s\right)\left)\right\le {M}^{\prime}$
for all
$s\in [0,t]$
and
${lim}_{l\to +\infty}f\left({x}_{{k}_{l}}\right(s),{\sigma}_{{k}_{l}}(s\left)\right)=f\left({x}^{*}\right(s),{\sigma}^{*}(s\left)\right)$
for almost all
$s\in [0,t]$
, applying the Lebesgue Dominated Convergence Theorem we have that
$${lim}_{l\to +\infty}{\int}_{0}^{t}f\left({x}_{{k}_{l}}\right(s),{\sigma}_{{k}_{l}}(s\left)\right)ds={\int}_{0}^{t}f\left({x}^{*}\right(s),{\sigma}^{*}(s\left)\right)ds,$$
and, consequently,
$${x}^{*}\left(t\right)={x}^{*}\left(0\right)+{\int}_{0}^{t}f\left({x}^{*}\right(s),{\sigma}^{*}(s\left)\right)ds.$$
It only remains to show that
$({x}^{*},{\sigma}^{*})$
belongs to
$\mathcal{T}\left[\chi \right]$
. Let
$t\ge 0$
so that
${lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left(t\right)={\sigma}^{*}\left(t\right)=j$
.
As
$\Gamma $
is finite, there exists
${l}^{*}>0$
such that
${\sigma}_{{k}_{l}}\left(t\right)=j$
for all
$l\ge {l}^{*}$
. Note that due to the definition of
$\mathcal{T}\left[\chi \right]$
, we also have that
${x}_{{k}_{l}}\left(t\right)\in {\chi}_{j}$
for all
$l\ge {l}^{*}$
. As
$\left\{{x}_{{k}_{l}}\right(t\left)\right\}$
converges to
${x}^{*}\left(t\right)$
and
${\chi}_{j}$
is closed, we deduce that
${x}^{*}\left(t\right)\in {\chi}_{j}$
. In consequence,
$$\begin{array}{ccc}{x}^{*}\left(t\right)\in {\chi}_{{\sigma}^{*}\left(t\right)}foralmostallt\ge 0.& & \end{array}$$ 
(4)

Now let
$t\ge 0$
be arbitrary. Due to ( 4 ) there exists a sequence
$\left\{{s}_{k}\right\}$
which converges to
$t$
so that
$t\le {s}_{k}$
,
${\sigma}^{*}\left({s}_{k}\right)={\sigma}^{*}\left(t\right)$
and
${x}^{*}\left({s}_{k}\right)\in {\chi}_{{\sigma}^{*}\left(t\right)}$
for all
$k$
. Then, from the continuity of
${x}^{*}$
and the fact that
${\chi}_{{\sigma}^{*}\left(t\right)}$
is closed, we have that
${lim}_{k\to \infty}{x}^{*}\left({s}_{k}\right)={x}^{*}\left(t\right)\in {\chi}_{{\sigma}^{*}\left(t\right)}$
and the proof is completed.
Lemma A.2
Assume that
$\chi $
is a closed covering of
${\mathbb{R}}^{n}$
and
$V:\Omega \times \Gamma \to \mathbb{R}$
, with
$\Omega $
an open subset of
${\mathbb{R}}^{n}$
, is continuous. Let
$\left\{\right({x}_{k},{\sigma}_{k}\left)\right\}$
be a sequence of maximal trajectories of ( 1 ) belonging to
${\mathcal{T}}_{V}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a[{\tau}_{D},{N}_{0}]$
,
${\tau}_{D}>0$
,
${N}_{0}\in \mathbb{N}$
, and suppose that there exists a compact subset
$B\subset \Omega $
such that
${x}_{k}\left(t\right)\in B$
for all
$t\ge 0$
and all
$k$
.
Then there exist a subsequence
$\left\{\right({x}_{{k}_{l}},{\sigma}_{{k}_{l}}\left)\right\}$
and a maximal trajectory
$({x}^{*},{\sigma}^{*})\in {\mathcal{T}}_{V}\cap \mathcal{T}\left[\chi \right]\cap \mathcal{T}a[{\tau}_{D},{N}_{0}]$
such that
$\left\{{x}_{{k}_{l}}\right\}$
converges to
${x}^{*}$
uniformly on compact sets of
${\mathbb{R}}_{+}$
and
${lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left(t\right)={\sigma}^{*}\left(t\right)$
a.e. on
${\mathbb{R}}_{+}$
.
$\square $
Proof. Since
$\left\{\right({x}_{k},{\sigma}_{k}\left)\right\}$
is uniformly bounded, from Proposition A.1 there exist a subsequence
$\left\{\right({x}_{{k}_{l}},{\sigma}_{{k}_{l}}\left)\right\}$
and a maximal trajectory
$({x}^{*},{\sigma}^{*})\in \mathcal{T}\left[\chi \right]\cap \mathcal{T}a[{\tau}_{D},{N}_{0}]$
such that
$\left\{{x}_{{k}_{l}}\right\}$
converges to
${x}^{*}$
uniformly on compact sets of
${\mathbb{R}}_{+}$
and
${lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left(t\right)={\sigma}^{*}\left(t\right)$
a.e. on
${\mathbb{R}}_{+}$
. Therefore the lemma follows provided
$({x}^{*},{\sigma}^{*})\in {\mathcal{T}}_{V}$
. As
${x}^{*}\left(t\right)\in B\subset \Omega $
for all
$t\in {\mathbb{R}}_{+}$
, it only remains to prove that for all pair of times
$t,{t}^{\prime}$
, with
$t<{t}^{\prime}$
and
${\sigma}^{*}\left(t\right)={\sigma}^{*}\left({t}^{\prime}\right)$
,
$V\left({x}^{*}\right(t),{\sigma}^{*}(t\left)\right)\ge V\left({x}^{*}\right({t}^{\prime}),{\sigma}^{*}({t}^{\prime}\left)\right)$
.
Let
$t,{t}^{\prime}$
be a pair of times such that
$t<{t}^{\prime}$
and
${\sigma}^{*}\left(t\right)={\sigma}^{*}\left({t}^{\prime}\right)=\gamma $
. Since
${lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left(t\right)={\sigma}^{*}\left(t\right)$
a.e.
on
${\mathbb{R}}_{+}$
, and
${\sigma}^{*}$
is piecewise constant and continuous from the right, there exists a pair of nonincreasing sequences
$\left\{{s}_{i}\right\}$
,
$\left\{{s}_{i}^{\prime}\right\}$
so that

∙
${s}_{i}<{s}_{i}^{\prime}$
for all
$i$
;

∙
${s}_{i}\searrow t$
and
${s}_{i}^{\prime}\searrow {t}^{\prime}$
;

∙
for every
$i$
,
${lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left({s}_{i}\right)={\sigma}^{*}\left({s}_{i}\right)=\gamma $
and
${lim}_{l\to \infty}{\sigma}_{{k}_{l}}\left({s}_{i}^{\prime}\right)={\sigma}^{*}\left({s}_{i}^{\prime}\right)=\gamma $
.
Fix
$i$
. Since
$\Gamma $
is finite, we have that
${\sigma}_{{k}_{l}}^{*}\left({s}_{i}\right)={\sigma}_{{k}_{l}}^{*}\left({s}_{i}^{\prime}\right)=\gamma $
for
$l$
large enough, say
$l\ge {l}^{*}$
. Then, from the definition of
${\mathcal{T}}_{V}$
and the fact that
${s}_{i}<{s}_{i}^{\prime}$
, it follows that
$V\left({x}_{{k}_{l}}\right({s}_{i}),{\sigma}_{{k}_{l}}({s}_{i}\left)\right)\ge V\left({x}_{{k}_{l}}\right({s}_{i}^{\prime}),{\sigma}_{{k}_{l}}({s}_{i}^{\prime}\left)\right)$
for all
$l\ge {l}^{*}$
. In consequence, by taking limit as
$l\to \infty $
we obtain that
$V\left({x}^{*}\right({s}_{i}),{\sigma}^{*}({s}_{i}\left)\right)\ge V\left({x}^{*}\right({s}_{i}^{\prime}),{\sigma}^{*}({s}_{i}^{\prime}\left)\right)$
and, a posteriori, letting
$i\to \infty $
we have that
$V\left({x}^{*}\right(t),{\sigma}^{*}(t\left)\right)\ge V\left({x}^{*}\right({t}^{\prime}),{\sigma}^{*}({t}^{\prime}\left)\right)$
.