Nonlinear Systems and Control Spring 2017

Chapter 2
Problem 1 - Solution
tan1 (ax1 ) x1 x2
 
>
Let := [a b c] and consider the vector field f (t, x, ) := .
bx21 cx2
The Jacobian matrices f (t,x,)
x
and f (t,x,)

are given by

 a   x1 
f (t, x, ) x2 x1
1+(ax1 )2
f (t, x, ) 1+(ax1 )2
0 0
= , = .
x 2bx1 c 0 x21 x2

Let 0 := [a0 b0 c0 ] be the nominal value of and denote by x(t, 0 ) the

unique solution of the nominal state equation

x = f (t, x, 0 ), x(t0 ) = x0 .

The sensitivity equation is given by

S(t) = A(t, 0 )S(t) + B(t, 0 ), S(t0 ) = 0,

where
   
f (t, x, ) f (t, x, )
A(t, 0 ) = , B(t, 0 ) = .
x x(t,0 ),0 x(t,0 ),0

One gets that

 1 
1+x1 (t,0 )2
x2 (t, 0 ) x1 (t, 0 )
A(t, 0 ) =
0 1

and " #
x1 (t,0 )
1+x1 (t,0 )2
0 0
B(t, 0 ) = .
0 x1 (t, 0 )2 x2 (t, 0 )
In this procedure, we need to solve first the nominal state equation and then
the linear time-varying sensitivity equation. This is usually done numerically.
Another equivalent approach is to solve these two equations simultaneously.
To this aim, we set  
x3 x5 x7
S=
x4 x6 x8

1 of 6
Nonlinear Systems and Control Spring 2017

and we consider the augmented equation

x1 = tan1 (x1 ) x1 x2
x2 = x2
 
1 x1
x3 = 2
x2 x3 x1 x4 +
1 + x1 1 + x21
x4 = x4
 
1
x5 = x2 x5 x1 x6
1 + x21
x6 = x6 + x21
 
1
x7 = x2 x7 x1 x8
1 + x21
x8 = x8 x2

with initial conditions x1 (t0 ) = x01 , x2 (t0 ) = x02 , x3 (t0 ) = x4 (t0 ) = x5 (t0 ) =
x6 (t0 ) = x7 (t0 ) = x8 (t0 ) = 0.

Problem 2 - Solution
1) The system is linear and in particular the vector field is globally Lipschitz.
By Theorem 2.3, this guarantees existence and uniqueness in [0, +).

2) Expanding in Taylor series the original vector field, one notices that

x 2 x3
ex 1 x =1 + x + + + 1 x
2 6
x2 x3
= + ( + ...).
2 | 6 {z }
0 for x0

Consider the initial condition x(0) = x0 , where x0 > 0. It is well known that
2
the differential equation x = x2 has finite escape time. Indeed the solution
to the Cauchy problem with initial condition (0, x0 ) has the form
2
x(t) = 2 , t 0.
x0
t

2 of 6
Nonlinear Systems and Control Spring 2017

x2
Therefore, limt 2 x(t) = +. Since ex 1 x 2
for x 0, the solution
x0
x
to the original differential equation x = e 1 x has a greater slope and so
escapes to infinity sooner. This proves that the differential equation 2) does
not admit a global solution in [0, +).

3) For x 6= 0 it is easy to show that the derivative of the vector field

f (x) = e|x| 1 is bounded
 
d 
 f (x) < 1 x =
6 0.
 dx 
The only thing that one needs to check is whether or not the derivative of f
is bounded for x = 0. One notices that
 +    
 d   d 
 dx f (0)  = 1,  dx f (0)  = 1.
   

Since in addition f is continuous, we deduce that f is globally Lipschitz.

Therefore, by Theorem 2.3 the solution exists and is unique for t [0, +).

4) We will first show that the o.d.e.

x1 (t) = x1 (t)3 , t0
x1 (0) = x01 ,
admits a global unique solution. Note that the vector field f (x) := x3
is not globally Lipschitz and so we cannot use Theorem 2.3. However, it
is locally Lipschitz at x, for every x R. Moreover, when x1 (t) > 0, the
derivative x1 (t) < 0 and when x1 (t) < 0, the derivative x1 (t) > 0. Thus
for any solution x1 () of the o.d.e. we have x1 (t) K, for all t 0, where
K := {x R : |x| x01 } is compact. Therefore, by Therem 2.6 we conclude
that the o.d.e. admits a unique global solution in [0, +). Finally, note that
the o.d.e.
x2 (t) = x1 (t) x2 (t) x1 (t)3 , t0
x2 (0) = x02 ,
admits a unique global solution in [0, ) as well, because the vector field
f (t, x) := x1 (t)xx1 (t)3 is continuous and globally Lipschitz in x uniformly
in t. Indeed f is continuously differentiable and fx
(t,x)
= 1 and so bounded.

3 of 6
Nonlinear Systems and Control Spring 2017

Problem 3 - Solution
Consider the o.d.e.
p
x(t) = x(t), t0
x(0) = 1.

Let x() be a solution of the system. Then, since x(t) 0, for all t 0, the
trajectory x() is monotonically increasing.
In particular, x(t) x(0) = 1, for
all t 0. We will show that f (x) := x, x 1 is globally Lipschitz. Indeed,
f is continuously differentiable and |f 0 (x)| 12 , for all x 1. Therefore by
Theorem 2.3 we conclude that the system admits a unique solution.
The answer is different if we consider the system
p
x(t) = x(t), t 0
x(0) = 0,

since f is not even locally Lipschitz for x = 0. This is because limx0+ f 0 (x) =
+. In this case, the assumptions of Theorem 2.3 are not satisfied and thus
we cannot conclude that the system admits a unique solution. In fact the
system has infinitely many solutions. Two different solutions are for example

t2
x(t) = 0 t 0 and x(t) = t 0 .
4
Moreover, it is easy to check that for any fixed c > 0, the trajectory
 (tc)2
, t [c, +)
x(t) = 4 ,
0, t [0, c)

is a solution as well.

Problem 4 - Solution
Since f1 is locally Lipschitz at x0 , there exist r1 > 0 and L1 > 0 such that

|f1 (x) f1 (y)| L1 |x y|, (1)

4 of 6
Nonlinear Systems and Control Spring 2017

for all x, y {z R : |z x0 | r1 }. Similarly, since f2 is locally Lipschitz

at x0 , there exist r2 > 0 and L2 > 0 such that

|f2 (x) f2 (y)| L2 |x y|, (2)

for all x, y {z R : |z x0 | r2 }. Set r := min{r1 , r2 }. Then (1) and (2)

hold for all x, y B(x0 , r) := {z R : |z x0 | r}.

1) It follows by the triangle inequality that

|(f1 + f2 )(x) (f1 + f2 )(y)| = |f1 (x) f1 (y) + f2 (x) f2 (y)|

|f1 (x) f1 (y)| + |f2 (x) f2 (y)| (L1 + L2 )|x y|,

for all x, y B(x0 , r). This proves that the sum is locally Lipschitz at x0 .

2) It follows using the triangle inequality that

|(f1 f2 )(x) (f1 f2 )(y)| = |f1 (x)f2 (x) f1 (y)f2 (y)| =

= |f1 (x)f2 (x) f1 (x)f2 (y) + f1 (x)f2 (y) f1 (y)f2 (y)|
|f1 (x)(f2 (x) f2 (y))| + |f2 (y)(f1 (x) f1 (y))|
|f1 (x)|L2 |x y| + |f2 (y)|L1 |x y| (|f1 (x)|L2 + |f2 (y)|L1 )|x y| ,
for all x, y B(x0 , r). Let M1 := supxB(x0 ,r) |f1 (x)| and M2 := supxB(x0 ,r) |f2 (x)|.
Then
|(f1 f2 )(x) (f1 f2 )(y)| (M1 L2 + M2 L1 )|x y| ,
for all x, y B(x0 , r). This proves that the product is locally Lipschitz at x0 .

3) Assume that f1 , f2 are globally Lipschitz functions with Lipschitz con-

stants L1 and L2 respectively. Then, using the definition of Lipschitz conti-
nuity, one gets that

|(f2 (f1 (x))) f2 (f1 (y))| L2 |f1 (x) f1 (y)| L2 L1 |x y| ,

for all x, y R. This proves that the composition is globally Lipschitz

continuous as well.

5 of 6
Nonlinear Systems and Control Spring 2017

Problem 5 - Solution

P is contractive [0, 1) : ||Ax Ay|| ||x y||, x, y Rn

[0, 1) : ||Av|| ||v||, v Rn .

Taking the square on both sides and rearranging one gets

||Av||2 2 ||v||2 v > A> Av 2 v > v v > (A> A 2 I)v 0.

So,

P is contractive [0, 1) : v > (A> A 2 I)v < 0, v Rn /{0}.

This holds if and only if all the eigenvalues of A> A 2 I are negative. In
particular one knows that if {1 , . . . , n } [0, +) are the eigenvalues of
A> A, then {1 2 , . . . , n 2 } are the eigenvalues of A> A 2 I. Thus,

P is contractive [0, 1) : i 2 , i = 1, . . . , n
i < 1, i = 1, . . . , n
p
i < 1, i = 1, . . . , n.

Thus P is contractive if and only if all the singular values of A are smaller
than 1. Moreover, when A> = A, if all the eigenvalues of A have absolute
value smaller than 1, then P is contractive.

6 of 6