AMES 207 Prof. M.

Krstic
Nonlinear Systems February 12, 1998

MIDTERM EXAM

Take home. Open books and notes.

Total points: 30
Due February 13, 1998, at 5:00 p.m. in Professor Krstic’s office.

Late submissions will not be accepted. Collaboration not allowed.

Problem 1.
Consider the system:

ẋ = −cx + y 2m x cos2 x
ẏ = −y 3

Using either Gronwall’s inequality or the comparison principle, show that

a) (5 pts) x(t) is bounded for all t ≥ 0 whenever c = 0 and m > 1.

b) (5 pts) x(t) → 0 as t → ∞ whenever c > 0 and m = 1.

Problem 2.
Consider the system:

ẋ = −x + yx + z cos x
ẏ = −x2
ż = −x cos x

a) (2 pts) Determine all the equilibria of the system.

b) (2 pts) Show that the equilibrium x = y = z = 0 is globally stable.

c) (2 pts) Show that x(t) → 0 as t → ∞.

d) (2 pts) Show that z(t) → 0 as t → ∞.

Problem 3.
With Chetaev’s theorem, show that the equilibrium at the origin of the following two systems
is unstable:

a)(4pts)

ẋ = x3 + xy 3
ẏ = −y + x2

b)(4pts)

ξ˙ = η + ξ 3 + 3ξη 2
η̇ = −ξ + η 3 + 3ηξ 2

Problem 4.
(4 pts) Calculate exactly (in closed form) the sensitivity function at λ0 = 0 for the system

ẋ = −x + tan−1 (λx).

How accurate is the approximation x(t, λ) ≈ x(t, λ0 ) + S(t)(λ − λ0 ) for large λ, say λ > 1?
AMES 207 Prof. M. Krstic
Nonlinear Systems Winter 1998

FINAL EXAM
Take home. Open books and notes.
Total points: 35
Due Saturday, March 14, 1998, at 12:00 noon in Professor Krstic’s office.

Late submissions will not be accepted. Collaboration not allowed.

1. (8 pts) Consider the system

ẋi = xi+1 − ci xi − ki si (x)xi + wi(x)d,

i = 1, . . . , n, xn+1 = 0, where ci , ki > 0 and |wi (x)| ≤ si (x). Show that the system is ISS
w.r.t. d. What is the type of the gain function (linear, quadratic, exponential, . . .)?

2. (9 pts) Using the center manifold theorem, determine whether the origin of the following
system is asymptotically stable:

ẋ1 = −x2 + x1 x3
ẋ2 = x1 + x2 x3
ẋ3 = −x3 − (x21 + x22 ) + x23

3. (9 pts) Consider the system

ẋ1 = x2  
ẋ2 = −x2 − sin ωt (x1 + sin ωt)2 + x3
1
ẋ3 = −xn3 − (x1 + sin ωt)2 + .
2

a) For n = 1, showthat
 for sufficiently large ω there exists an exponentially stable periodic
1
orbit in an O ω − neighborhood of the origin.

b) What can you claim for n = 3?

4. (9 pts) Show that, for sufficiently small ε, the origin of the system
ẋ = x2 + z + cos (εy) − 1
εẏ = −y + x2 − x
ε2 ż = −z + sin y + εx3

is exponentially stable. (Hint: treat µ = ε2 as a separate small parameter.) Since the

system has three (rather than two) time scales, it has three levels of invariant manifolds –
slow, medium, and fast. Without going into high accuracy, give the approximate expressions
for these manifolds and discuss the trajectories of the system.
MAE 281A Prof. M. Krstic
Nonlinear Systems February 15, 2000

MIDTERM EXAM

Take home. Open books and notes.

Total points: 25
Due February 16, 2000, at 5:00 p.m. in Professor Krstic’s office.

Late submissions will not be accepted. Collaboration not allowed.

Problem 1.
(6 pts) Let g, h, and y be three positive functions on (0, ∞) such that
Z ∞
g(t)dt ≤ C1
0
Z ∞
eδt h(t)dt ≤ C2
0
Z ∞
eδt y(t)dt ≤ C3 ,
0

where δ, C1 , C2 , C3 are positive constants. Assuming that

ẏ ≤ g(t)y + h(t) , ∀t ≥ 0

using Gronwall’s lemma show that

y(t) ≤ [C2 + δC3 + y(0)]eC1 −δt .

Problem 2.
(6 pts) Consider the system:

ẋ = A(x, y)x + B(x)y

ẏ = −GB(x)T x ,

where x(t), y(t) are vectors of arbitrary dimensions, A(x, y) is a matrix valued function that
satisfies
A(x, y) + A(x, y)T ≤ −qI , q>0
and G is a positive definite symmetric matrix. Show that the equilibrium x = 0, y = 0 is
globally stable, that x(t) converges to zero, and that y(t) converges to the null space of B(0).
If you can’t solve the problem for general G, solve it for G = I to receive partial credit.
Problem 3.
(6 pts) With Chetaev’s theorem, show that the equilibrium at the origin of the following
system is unstable:
q
ẋ = |x|x + xy |y|
q
ẏ = −y + |x| |y| .

Don’t worry about uniqueness of solutions.

Problem 4.
(7 pts) Calculate exactly (in closed form) the sensitivity function at λ0 = 1 for the system

ẋ = −λx3 .

What is the approximation x(t, λ) ≈ x(t, λ0 ) + S(t)(λ − λ0 ) for λ = 7/2?

MAE 281A Prof. M. Krstic
Nonlinear Systems Winter 2001

FINAL EXAM
Take home. Open books and notes.
Total points: 50
Due Thursday, March 22, 2001, at 1:00 pm in Professor Krstic’s office.

Late submissions will not be accepted. Collaboration not allowed.

1. (9 pts) Using the Lyapunov function candidate V = 12 (x2 + y 2 + z 2 ), study stability of the origin of
the system
ẋ = −x + x2 z
ẏ = z
ż = −y − z − x3 .

2. (9 pts) Show that the following system is ISS

ẋ = −x + x1/3 y + p2
ẏ = −y − x4/3 + p3
ṗ = −p + u .

3. (8 pts) Using the center manifold theorem, determine whether the origin of the following system is
asymptotically stable:
ẏ = yz + 2y 3
ż = −z − 2y 2 − 4y 4 − 2y 2 z .

4. (8 pts) Using averaging theory, analyze the following system:

ẋ = ǫ[−x + 1 − 2(y + sin t)2 ]
ẏ = ǫz
1
  
ż = ǫ −z − sin t x + (y + sin t)2 .
2

5. (8 pts) Using singular perturbation theory, study local exponential stability of the origin of the system
ẋ = y
ẏ = −z
ǫż = −z + sin x + y .
Is the origin globally exponentially stable?
1
6. (8 pts) Consider the feedback system with a linear block (like in class) and a non-
s(s + 1)(s + 2)
linearity sgn(y) + |y|y (note that it is an odd nonlinearity, so the describing functions method applies,
and note that the first term was already studied in class). First, find the describing function for the
nonlinearity. Then, determine if the feedback system is likely to have any periodic solutions.
MAE 281A Prof. M. Krstic
Nonlinear Systems Winter 2002

FINAL EXAM

Take home. Open books and notes.

Total points: 65
Due Friday, March 22, 2002, at 4:00 pm in Professor Krstic’s office.

Late submissions will not be accepted. Collaboration not allowed.

Each problem is worth 13 points

Problem 1. Consider the system

ẋ = −x + yx sin x
ẏ = −y + zy sin y
ż = −z .

Using Gronwall’s lemma (twice), show that

|z0 |
|x(t)| ≤ |x0 |e|y0 |e e−t , ∀t ≥ 0.

Problem 2. Analyze uniform stability of the origin of the linear time-varying system

ẋ = y
ẏ = −y − (2 + sin t)x

using the Lyapunov function

y2
V = x2 + .
2 + sin t
Does your analysis guarantee that y(t) → 0 as t → ∞.

Problem 3. Using the Lyapunov function candidate

x4 y 2 z 4
V = + + ,
4 2 4
study stability of the origin of the system

ẋ = y
ẏ = −x3 − y 3 − z 3
ż = −z + y .
Problem 4. Using the Chetaev function
V = xz
prove that the origin of the system

ẋ = yz + az
ẏ = −xz
ż = xy + ax,

where a > 0 is a constant, is unstable. (This problem is related to instability of rigid body spinning
motion around the “intermediate” axis.)

Problem 5. Show that the ISS gain function of the system

1
ẋ = (3 + cos(u))sgn(x) log +y
1 + |x|
x2
ẏ = −(2 + x2 )|y|y + u
1 + x2
from u to x is √
r
γ(r) = e − 1.
Hint: Use Lyapunov functions of the form V1 (x) = |x| and V2 (y) = |y|.
MAE 281A Prof. M. Krstic
Nonlinear Systems Winter 2003

FINAL EXAM

Take home. Open books and notes.

Total points: 65
Due Friday, March 14, 2003, at 4:00 pm in Professor Krstic’s office. (4pm is a hard deadline, I am leaving
the office at 4pm.)

Late submissions will not be accepted. Collaboration not allowed.

Problem 1. Consider the system

! "
ṗ = −µ!γ sin φ r 2 sin2 t − q
! "
q̇ = µ!γ r 2 sin2 t − q
φ̇ = µ!
! "
ṙ = µr cos2 t 1 + (p + sin φ)2 − r 2 sin2 t ,

where r(0) ≥ 0 (note that this implies that r(t) ≥ 0 for all time because r = 0 sets ṙ = 0). Denote
   
p 0
x= q  and xr (t) =  ) 3
   

r 2 1 + sin(µ!t)2

Show that, for sufficiently small µ, !, and γ, the solution x(t, µ, !, γ) locally exponentially converges to
xr (t) + O(µ + ! + γ), at least on a finite time interval.
Hint. This is a complicated problem that involves four time scales. (I hope it will not take you as much
time to solve it as it took me to construct it and double check its solvability.) The four time scales are
(going from fastest to slowest):

• sin t

• r

• sin(µ!t)

• p, q.

Your analysis should apply the following steps:

• One step of averaging for the complete system, treating µ as small.

• One step of singular perturbation, treating r as fast and ! as small, and introducing the new time
τ = !t to put the system (after averaging) into the standard singular perturbation form. To derive
the boundary layer model, you will need to introduce “another” time variable t̂ = τ /! = t.
 • A second step of averaging on the (p, q) system with φ = µτ = µ!t as time, treating γ as small
(again).

Note that the hardest part of the problem is not to mechanically perform the approximations but to
connect them all, through appropriate theorems, to draw the final conclusion. Make sure you do quote
the theorems as you go from the last step of simplification backwards towards the original system. I will
be quite unimpressed to see that you only know how to calculte an average system or how to find a quasi
steady state. Note that, in order to draw the final conclusion, exponential stability needs to be satisfied
every step of the way.

Problem 2. Consider the system

ẋ = −x + xz + y(1 − y)
ẏ = −x(1 − y)
ż = −x2 .

Give the most precise statement you can on stability and convergence (global and local) of solutions of
this system. Note that this is an open ended problem. Since the system has two entire lines of equilibria,
analyzing all of them might take many days of work, leading you to use not only the Lyapunov, LaSalle
invariance, and linearization theorems, but even the center manifold and Chetaev theorems. Go as far as
you can with your ideas and these tools.
Hints. If you try to study individual equilibria, the first thing to note is that, since they all belong to
continuous sets of equilibria, none of them can be asymptotically stable. So, the equilibria fall into one
of the two categories: stable or unstable. By taking linearizations around equilibria, you will note that
all of them have at least one eigenvalue at zero in their Jacobians. So, unless you find them unstable
by linearization, you may need the center manifold theorem. Note that, since none of the equilibria are
asymptotically stable, the center manifold theorem should work only for the equilibria that happen to be
unstable. Don’t immediately look for complicated center manifolds—trivial ones (h(·) = 0) will carry you
a long way. For those equilibria that happen to be stable you can use Lyapunov functions parametrized
by the equilibria. For one of the equilibria, (0,1,1), even center manifold is not enough and the stability
question needs to be resolved by direct Lyapunov or Chetaev. I personally have not figured out this one
as of this writing.
MAE 281A Prof. M. Krstic
Nonlinear Systems Winter 2004

FINAL EXAM
Take home. Open books and notes.
Total points: 65
Due Wednesday, March 17, 2004, at 1:00 pm in Professor Krstic’s office.

Late submissions will not be accepted. Collaboration not allowed.

1. (13 pts) Using the Lyapunov function candidate V = 21 (x2 + y 2 + z 2 ), study stability of the origin of
the system

ẋ = −x + x2 z
ẏ = z
ż = −y − z − x3 .

2. (13 pts) Show that the following system is ISS

ẋ = −x + x1/3 y + p2
ẏ = −y − x4/3 + p3
ṗ = −p + u .

3. (13 pts) Using averaging theory, analyze the following system:

ẋ = [−x + 1 − 2(y + sin t)2 ]
ẏ = z
1
  
ż =  −z − sin t x + (y + sin t)2 .
2

4. (13 pts) Using singular perturbation theory, study local exponential stability of the origin of the
system

ẋ = y
ẏ = −z
ż = −z + sin x + y .

Is the origin globally exponentially stable?

1
5. (13 pts) Consider the feedback system with a linear block (like in class) and a
s(s + 1)(s + 2)
nonlinearity sgn(y) + |y|y (note that it is an odd nonlinearity, so the describing functions method applies,
and note that the first term was already studied in class). First, find the describing function for the
nonlinearity. Then, determine if the feedback system is likely to have any periodic solutions.
MAE 281A Prof. M. Krstic
Nonlinear Systems Winter 2006

FINAL EXAM

Take home. Open books and notes.

Total points: 65
Due Tuesday, March 21, 2006, at 5:00 pm in Professor Krstic’s office.

Collaboration not allowed.

1. (20 pts) Show that the following system has an unstable equilibrium at the origin:

ẋ1 = x31 + 2x32

ẋ2 = x1 x22 + x32 .

Hint: For second order systems one should always first try Chetaev functions of the form V = x21 − ax22 ,
where a is some positive constant which you are free to choose in the analysis.

2. (25 pts) Show that the following system is ISS

ẋ = −x + y 3
ẏ = −y − f (x) + p2
ṗ = −p + u ,

where f (x) is a function such that 0x f (ξ)dξ and xf (x) are positive definite and radially unbounded
R

functions.
[An example of such a function is the function f (x) = ex − 1—convince yourself that this is so. A
simpler, trivial example is the function is f (x) = x.]
This is the hardest problem on the exam. You can solve it either by showing that the (x, y)-system
is ISS with respect to p and by noting that p is ISS with respect to u, or by directly building an ISS-
Lyapunov function for the whole (x, y, p)-system (the latter is more difficult but more impressive if you
can
Rx
do it). In any case you will have to construct some Lyapunov functions. Hints: Use the terms
4
0 f (ξ)dξ and y /4 in those Lyapunov functions. In showing that the (x, y)-system is ISS with respect
to p you should use Young’s inequality (the version with powers of 4 and 4/3).

3. (20 pts) Using averaging theory, analyze the following system:

ẋ1 = [(x1 − sin ωt)2 − x2 ] sin ωt

ẋ2 = −(x1 − sin ωt)2 − x2 .
MAE 281A Prof. M. Krstic
Nonlinear Systems Winter 2007

FINAL EXAM

Take home. Open books and notes.

Total points: 75
Tuesday, March 20, 2006

Collaboration not allowed.

1. (10 pts) Prove global stability of the origin of the system

ẋ1 = x2
x1
ẋ2 = − .
1 + x22

2. (10 pts) Prove global asymptotic stability of the origin of the system

ẋ1 = −x32
ẋ2 = x1 − x2 .

Is the origin exponentially stable (at least locally)?

3. (10 pts) Which of the state variables of the following system are guaranteed to converge to zero from
any initial condition?

ẋ1 = x2 + x1 x3
ẋ2 = −x1 − x2 + x2 x3
ẋ3 = −x21 − x22 .

4. (10 pts) Using averaging theory, analyze the behavior of the following system for large ω, for both
a = 1 and a = −1:

ẋ1 = − sin x1 + 2x2 + (x1 + 4x2 sin ωt) sin ωt

 
ẋ2 = −2x1 cos ωt + x22 sin ωt cos ωt − ax2 .

5. (10 pts) Consider the following control system:

ẋ = A11 x + A12 z + B1 u
εż = A21 x + A22 z .

Assume that the matrix A22 is Hurwitz and that there exists a matrix/vector K (of appropriate dimen-
sions) such that
A11 − A12 A−1
22 A21 + B1 K
is also Hurwitz. Prove that the “partial-state” feedback law

u = Kx

exponentially stabilizes the equilibrium (x, z) = (0, 0) for sufficiently small ε.

6. (5 pts) Show that the following system is ISS and determine its gain function:

ẋ = −x3 + xu .

7. (5 pts) Show that the following system is ISS and determine its gain function:

ẋ = −x + u3 .

8. (5 pts) Show that the following system is ISS and guess its gain function:

ẋ = −x3 + xy
ẏ = −y + u3 .
MAE 281A Prof. M. Krstic
Nonlinear Systems Winter 2008

FINAL EXAM
Open book and class notes. Collaboration not allowed.
Total points: 65

1. (10 pts) Prove global asymptotic stability of the origin of the system
 
ẋ1 = x2 − 2x21 + x22 x1
 
ẋ2 = −x1 − 2 2x21 + x22 x2 .

Is the origin locally exponentially stable and why or why not?

2. (10 pts) Which of the states of the following system are guaranteed to go to zero (provide a proof for
your answer):

ẋ = z 2
ẏ = z cos z
ż = −z − zx − y 3 cos z .

3. (10 pts) Find the power-two (quadratic) approximation of the center manifold of the following system:

ẏ1 = y1 y22 + y1 z
ẏ2 = −y2 y12 + y2 z
ż = −z + y12 − y22 .

4. (10 pts) Using the center manifold theorem and the suitable Lyapunov or Chetaev theorem, determine
asymptotic stability or instability of the origin of the system in Problem 3.

5. (15 pts) Analyze the following system using the method of averaging for large ω:

ẋ1 = (x2 sin ωt − 2) x1 − x3

 
ẋ2 = −x2 + x22 sin ωt − 2x3 cos ωt cos ωt
ẋ3 = 2x2 − sin x3 + (4x2 sin ωt + x3 ) sin ωt .

6. (10 pts) Consider the system

ẋ = −x3 + u .
Show that the following holds
Z ∞ Z ∞ 2
4
x(τ ) dτ ≤ u(τ )4/3 dτ + x(0)2 .
0 0 3
Explanation for further learning (not needed to solve the problem): This property is called integral ISS.
Recall that we proved the standard ISS property in class for the same example. Recall also the discussion
of input-output properties of linear systems in class. For the linear example ẋ = −x + u we have input-
output stability in both L2 and L∞ norms, with gains equal to 1 in both cases. While the standard ISS
property is the nonlinear analog of L∞ stability, the ‘integral ISS’ is the nonlinear analog of L2 stability.
The above inequality says that the integral ISS gain function of the system ẋ = −x3 + u is the the cube
√
root function, 3 · (when x(t) is penalized in the sense of the L4 norm). This statement is the nonlinear
analog of the statement that the transfer function 1/(s + 1) has an H∞ norm of 1, i.e., that the linear
example ẋ = −x + u has an L2 induced gain of 1.
MAE 281A Prof. M. Krstic
Nonlinear Systems Winter 2009

FINAL EXAM
Open book and class notes. Collaboration not allowed.
Total points: 65

The first four problems in this exam are dedicated to “Chaplygin’s sleigh” (1895):

The nominal/ideal model for this system is given by

ẋ = v cos θ (1)
ẏ = v sin θ (2)
θ̇ = ω (3)

v̇ = aω 2 (4)
ma
ω̇ = − vω , (5)
I + ma2
where (x, y) is the position of the “knife edge” of the sleigh, θ is the heading angle, v is the surge velocity of the
sleigh, ω is the angular velocity, a is the distance from the knife edge to the sleigh’s center of mass, m is the
mass of the sleigh, and I is the moment of interia of the sleigh.
For those of you familiar with the ‘nonholonomic unicycle,’ note that the first three equations of the Chaplygin
sleigh model (1)–(3) are the same as the unicycle model. Those are the kinematic equations. We are actually not
going to use them in this exam, but I provide them to you so that you can physically appreciate the motivation
for the problems.
The remaining two equations of the Chaplygin sleigh (4), (5) are the basis of Problems 1–4 of this exam.
To physically imagine the motion of the Chaplygin sleigh, note that the knife edge is not able to slip sideways
(like an ice skate) but it can slide forward. Note also that the sliding posts are there only to provide support
and impose no friction, so the sleigh can freely rotate around the knife edge as the center of rotation (while at
the same time sliding).
Observe that the crucial quantities in the entire problem are the initial conditions for the velocity v(0) and
the angular velocity ω(0). Based on those quantities, the sleigh may go forward (though not straight) while
spinning around, creating complex figures in the (x, y) space.

Problem 1. (12 pts) What are the trajectories of the Chaplygin sleigh model in the (v, ω) space? Note that I
am not asking you to find the solutions (v(t), ω(t)) as functions of time but only the trajectory curves (phase
portrait) in the (v, ω) plane. Note also that this is actually a problem of finding a Lyapunov function for the
(v, ω) system.
Problem 2. (6 pts) Suppose now that the posts are subject to friction in the rotational (yaw) degree of freedom,
namely, that the (v, ω) equations are given by

v̇ = aω 2 (6)
ma
ω̇ = −hω − vω , (7)
I + ma2
where h > 0 is the friction coefficient. What are the (v, ω) trajectories in this case?

Problem 3. (10 pts) Suppose now that the blade itself is subject to friction in the surge degree of freedom,
namely, that the (v, ω) equations are given by

v̇ = −gv + aω 2 (8)
ma
ω̇ = − vω , (9)
I + ma2
where g > 0 is the friction coefficient. What type of stability property holds for the equilibrium v = ω = 0 and what
is the physical meaning of this? Is the origin locally exponentially stable?

Problem 4. (12 pts) Now, as the final problem on the Chaplygin sleigh, suppose that friction exists in both
the surge and yaw degrees of freedom (h, g > 0), and, more importantly, that the knife edge is allowed to slip
sideways slightly, namely, that the friction coefficient on the knife edge in the sideways direction is not infinite
but finite. In this case the (v, ω) equations are replaced by the three-equation model (v, ω, σ) given by

v̇ = −gv + aω 2 + ωσ (10)

ω̇ = −hω + σ (11)
I + ma2
σ̇ = − σ − vω , (12)
ma
where ε > 0 is a small parameter which is inversely proportional to the friction coefficient on the sideways motion
of the knife edge. Note that σ̇ is proportional to the derivative of the angular acceleration and, as such, can be
referred to as rotational “jerk.” Prove local exponential stability of the equilibrium v = ω = σ = 0 for small  > 0
(and g, h > 0) by using the singular perturbation approach. What can you state in the case g = h = 0 using Theorem
11.1 in Khalil?

Problem 5. (10 pts) Consider the system

 
ẋ = −x − sin ωt (z + sin ωt)2 + y (13)
1
ẏ = −y − (z + sin ωt)2 + (14)
2
ż = x . (15)
 
1
Show that for sufficiently large ω there exists an exponentially stable periodic orbit in an O ω neighborhood of the
origin.

Problem 6. (15 pts) Consider the system

ẋ = −x + y 3 (16)
x
ẏ = −y − √ + z2 (17)
1+x2

ż = −z + u . (18)

Show that this system is ISS using the Lyapunov function

p y4 z8
V = 1 + x2 − 1 + + (19)
4 2