Information Technology Reference
In-Depth Information
H
A
( I + X 1 )
,
X 2 .
Substituting these approximations into (8.10) and discarding all terms quadratic or
higher order in ( X 1
,
X 2 ) yields
X 1
X 2
=
X 1
X 2
k H I 3 I 3
.
(8.28)
k A I 3 0
Since k H ,
0, the linearized error system is exponentially stable. This proves the
local exponential stability of the equilibrium ( I
k A >
,
0).
Lastly, let us prove part iii) . First, we remark that the function V is constant and
strictly positive on the set E u . This can be easily verified from (8.8) and the defini-
tion of E u , using the fact that on this set A = 0, H T
H = H 2 =
2 I +( 1
λ
) vv T ,and
λ
λ
2
tr( vv T )=1since
= 1. We denote by V u the value of V on E u . The fact that V u is
strictly positive readily implies (in accordance with part ii) )that E s is an asymptoti-
cally stable equilibrium, since V is nonincreasing along the system's solutions, and
each of them converges to E s
v
E u . Using the same arguments, the proof of part iii)
reduces to showing that for any point ( H 0 ,
0)
E u , and any neighborhood
U
of this
point, one can find ( H 1 , A 1 )
∈U
such that
V ( H 1 , A 1 )
<
V u .
(8.29)
˙
Let H (
H = HC with C a constant
.
) denote a smooth curve on SL (3), solution of
(3) that will be specified latter on. We also assume that ( H (0)
element of
sl
,
0)
E u .
H ( t )
2
Let f ( t )=
I
/
2 so that, by (8.8), f (0)= V u . The first derivative of f is
given by
H ( t )) T ˙
f ( t )=
H ( t ))
tr(( I
H ( t )) T
H ( t ) C )
=
tr(( I
( H T ( t )( I
H ( t )))
=
P
,
C
.
For all elements ( H 0 ,
( H 0 ( I
H 0 )) = 0, so that f (0)=0. We now
0)
E u , one has
P
calculate the second order derivative of f :
f ( t )=tr( ˙
H ( t ) T ˙
H ( t )) T ¨
H ( t ))
H ( t ))
tr(( I
tr ( I
H ( t ) C
= tr( ˙
H ( t ) T ˙
H ( t )) T ˙
H ( t ))
where we have used the fact that C is constant. Evaluating the above expression at
t = 0 and replacing
˙
H (0) by its value H (0) C yields
tr ( I
H (0) C 2
f (0)=
H (0) C
H (0)) T
2
.
(8.30)
Search WWH ::




Custom Search