tensorkelechi/arxiv-cstext · Datasets at Hugging Face

text stringlengths 0 89.3k
where X1X
1RnnX2Rnm and X3X
3Rmm
andXstrictly copositive 23 The third one is the piecewise
quadratic function as in 24 that is for any xRn
VPWQxxP1xifxC1
xP2xifxC213
where P1P
1RnnandP2P
2Rnn This function
is also a conewise quadratic function with the same partition
as 9
The relationship between these three Lyapunov function
candidates unfolds as follows Function Vis a particular in
stance of Vquad Specifically if we set in 12 X1APAQ
X2APBS and X3BPBR then we obtain Vquad
V Function VPWQ is the most general form encompassing
both 11 and 12 as it accommodates Vquadwhich is more
general than V by setting P1X1KX
2X2KKX3K
andP2X1 leading to VPWQVquad Having identified
the most general Lyapunov function candidate among these
three options for system 9 LMI conditions under which the
stability of the origin of 9 would be guaranteed using 13
are provided in Appendix B Regrettably these conditions are
infeasible for the system discussed in Section IIB Further
more all our attempts to ensure the stability of this system
1This can be seen by noticing that the necessary and sufficient optimality
KarushKuhnTucker KKT conditions of the quadratic problem 5 write
RBPBvRBPBKxtλ0v0λ0λv0 where
λRis a Lagrange multiplier As a result the closedloop system 15 can
also be written as a discretetime LCS xt1AxtButytCxtDut
0utyt0 where CRBPBKandDRBPBbased on other existing tools from the literature 3 5 13
16 failed
Now for all the tested values of QRandSin 2 all
the solutions of the corresponding system 9 have the dis
tinctive feature to switch a maximum of 4times which once
analytically established allows concluding on the stability of
the origin for the considered system see Section VIB In the
following we formalize these findings for general CLS
D Problem statement
Motivated by the above developments we consider in this
paper general discretetime CLS of the form
xt1Aixt xtCi i1 m 14
where mNis the number of nonempty closed convex
conesCii1 m which partition the state space Rn
ie
C1 CmRn
intCiintCjfor all ij1 m
As in 3 we require that AiξAjξfor any ξCiCj This
implies the uniqueness of solutions for system 14 for any
initial condition For the sake of convenience and with some
slight abuse of notation we use ϕto denote solutions to 14
so that for initial condition xRn the corresponding solution
to 14 at time tNis denoted ϕt x
We refer to the dynamics of 14 for a given i1 m
as a mode We next formalize what is meant by a switch and
the maximum number of switches
Definition 1 Switch Given initial condition xRn solution
ϕ xto 14 switches at time tNif for any i1 m
such that ϕt1 xCiϕt xCi
A switch of a solution means that the solution has left the
cone it lies at the last time step to enter a different cone
In view of Definition 1 the set of switching instants for the
solution initialized at xRnis
JxtNi1 mst
ϕt1 xCiandϕt xCi15
Thus we can define the number of switches exhibited by a
solution to 14
Definition 2 Number of switches Consider system 14 and
xRn the number of switches exhibited by the solution
ϕ xisςxcardJx
In view of Definition 1 a solution ϕ xfor some xRn
exhibits a finite number of switches if and only if ςx
The main objective of this work is to derive conditions under
which all solutions to system 14 exhibits a finite number of
switches see Section IV This is motivated by the fact that the
stability analysis simplifies for this class of CLS as we show
in the next sectionDAAFOUZ et al CONEWISE LINEAR SYSTEMS WITH FINITEL Y MANY SWITCHES 5
III S TABILITY RESULTS
We first provide a necessary and sufficient condition for
x0to be GES for system 14 again assuming all its
solutions switch a finite number of times conditions to ensure
this property are provided in Section IV We then show that
this stability property is robust in the sense that an input
tostate stability property holds when 14 is perturbed by
exogenous disturbances
A Global exponential stability
We define next the set F which characterizes the region of
the state space where solutions to 14 stop switching
FxRn i1 mtNAt
ixCi16
In other words FxRnςx0with the notation of
Definition 2 Obviously 0FandFis forward invariant for
system 14 The next theorem gives a necessary and sufficient
condition for the origin of system 14 to be GES ie there
exist c11andc20such that for any xRnϕt x
c1ec2txfor any tN when we know that any solution
exhibit a finite number of switches Again conditions to ensure
the latter property are provided in Section IV

where X1X

1RnnX2Rnm and X3X

3Rmm

andXstrictly copositive 23 The third one is the piecewise

quadratic function as in 24 that is for any xRn

VPWQxxP1xifxC1

xP2xifxC213

where P1P

1RnnandP2P

2Rnn This function

is also a conewise quadratic function with the same partition

as 9

The relationship between these three Lyapunov function

candidates unfolds as follows Function Vis a particular in

stance of Vquad Specifically if we set in 12 X1APAQ

X2APBS and X3BPBR then we obtain Vquad

V Function VPWQ is the most general form encompassing

both 11 and 12 as it accommodates Vquadwhich is more

general than V by setting P1X1KX

2X2KKX3K

andP2X1 leading to VPWQVquad Having identified

the most general Lyapunov function candidate among these

three options for system 9 LMI conditions under which the

stability of the origin of 9 would be guaranteed using 13

are provided in Appendix B Regrettably these conditions are

infeasible for the system discussed in Section IIB Further

more all our attempts to ensure the stability of this system

1This can be seen by noticing that the necessary and sufficient optimality

KarushKuhnTucker KKT conditions of the quadratic problem 5 write

RBPBvRBPBKxtλ0v0λ0λv0 where

λRis a Lagrange multiplier As a result the closedloop system 15 can

also be written as a discretetime LCS xt1AxtButytCxtDut

0utyt0 where CRBPBKandDRBPBbased on other existing tools from the literature 3 5 13

16 failed

Now for all the tested values of QRandSin 2 all

the solutions of the corresponding system 9 have the dis

tinctive feature to switch a maximum of 4times which once

analytically established allows concluding on the stability of

the origin for the considered system see Section VIB In the

following we formalize these findings for general CLS

D Problem statement

Motivated by the above developments we consider in this

paper general discretetime CLS of the form

xt1Aixt xtCi i1 m 14

where mNis the number of nonempty closed convex

conesCii1 m which partition the state space Rn

ie

C1 CmRn

intCiintCjfor all ij1 m

As in 3 we require that AiξAjξfor any ξCiCj This

implies the uniqueness of solutions for system 14 for any

initial condition For the sake of convenience and with some

slight abuse of notation we use ϕto denote solutions to 14

so that for initial condition xRn the corresponding solution

to 14 at time tNis denoted ϕt x

We refer to the dynamics of 14 for a given i1 m

as a mode We next formalize what is meant by a switch and

the maximum number of switches

Definition 1 Switch Given initial condition xRn solution

ϕ xto 14 switches at time tNif for any i1 m

such that ϕt1 xCiϕt xCi

A switch of a solution means that the solution has left the

cone it lies at the last time step to enter a different cone

In view of Definition 1 the set of switching instants for the

solution initialized at xRnis

JxtNi1 mst

ϕt1 xCiandϕt xCi15

Thus we can define the number of switches exhibited by a

solution to 14

Definition 2 Number of switches Consider system 14 and

xRn the number of switches exhibited by the solution

ϕ xisςxcardJx

In view of Definition 1 a solution ϕ xfor some xRn

exhibits a finite number of switches if and only if ςx

The main objective of this work is to derive conditions under

which all solutions to system 14 exhibits a finite number of

switches see Section IV This is motivated by the fact that the

stability analysis simplifies for this class of CLS as we show

in the next sectionDAAFOUZ et al CONEWISE LINEAR SYSTEMS WITH FINITEL Y MANY SWITCHES 5

III S TABILITY RESULTS

We first provide a necessary and sufficient condition for

x0to be GES for system 14 again assuming all its

solutions switch a finite number of times conditions to ensure

this property are provided in Section IV We then show that

this stability property is robust in the sense that an input

tostate stability property holds when 14 is perturbed by

exogenous disturbances

A Global exponential stability

We define next the set F which characterizes the region of

the state space where solutions to 14 stop switching

FxRn i1 mtNAt

ixCi16

In other words FxRnςx0with the notation of

Definition 2 Obviously 0FandFis forward invariant for

system 14 The next theorem gives a necessary and sufficient

condition for the origin of system 14 to be GES ie there

exist c11andc20such that for any xRnϕt x

c1ec2txfor any tN when we know that any solution

exhibit a finite number of switches Again conditions to ensure

the latter property are provided in Section IV