Introduction to Nonlinear Control

Lecture # 2

Time-Invariant Systems

– p. 1/5
Lyapunov’s Method
Let V (x) be a continuously differentiable function defined in
a domain D ⊂ Rn ; 0 ∈ D . The derivative of V along the
trajectories of ẋ = f (x) is
n n
X ∂V X ∂V
V̇ (x) = ẋi = fi (x)
i=1
∂xi i=1
∂xi
 
f1 (x)
h i
 f2 (x)


∂V ∂V ∂V
= ∂x1
, ∂x2
, ... , ∂xn

 .. 

 . 
fn (x)
∂V
= f (x)
∂x
– p. 2/5
If φ(t; x) is the solution of ẋ = f (x) that starts at initial
state x at time t = 0, then

d 
V̇ (x) = V (φ(t; x))
dt t=0

If V̇ (x) is negative, V will decrease along the solution of

ẋ = f (x)

If V̇ (x) is positive, V will increase along the solution of

ẋ = f (x)

– p. 3/5
Lyapunov’s Theorem:
If there is V (x) such that

V (0) = 0 and V (x) > 0, ∀ x ∈ D/{0}

V̇ (x) ≤ 0, ∀x∈D
then the origin is stable

Moreover, if

V̇ (x) < 0, ∀ x ∈ D/{0}

then the origin is asymptotically stable

– p. 4/5
 Furthermore, if V (x) > 0, ∀ x 6= 0,

kxk → ∞ ⇒ V (x) → ∞

and V̇ (x) < 0, ∀ x 6= 0, then the origin is globally

asymptotically stable

Proof: D 0 < r ≤ ε, Br = {kxk ≤ r}

B
r
α = min V (x) > 0
kxk=r
Bδ
Ωβ 0<β<α
Ωβ = {x ∈ Br | V (x) ≤ β}
kxk ≤ δ ⇒ V (x) < β

– p. 5/5
Solutions starting in Ωβ stay in Ωβ because V̇ (x) ≤ 0 in Ωβ

x(0) ∈ Bδ ⇒ x(0) ∈ Ωβ ⇒ x(t) ∈ Ωβ ⇒ x(t) ∈ Br

kx(0)k < δ ⇒ kx(t)k < r ≤ ε, ∀ t ≥ 0

⇒ The origin is stable
Now suppose V̇ (x) < 0 ∀ x ∈ D/{0}. V (x(t) is
monotonically decreasing and V (x(t)) ≥ 0

lim V (x(t)) = c ≥ 0
t→∞

lim V (x(t)) = c ≥ 0 Show that c = 0

t→∞

Suppose c > 0. By continuity of V (x), there is d > 0 such

that Bd ⊂ Ωc . Then, x(t) lies outside Bd for all t ≥ 0

– p. 6/5
 ⇒ V̇ (x(t)) ≤ −γ ∀ t ≥ 0, for some γ > 0
Z t
V (x(t)) = V (x(0)) + V̇ (x(τ )) dτ ≤ V (x(0)) − γt
0
This inequality contradicts the assumption c > 0
⇒ The origin is asymptotically stable

The condition kxk → ∞ ⇒ V (x) → ∞ implies that the

set Ωc = {x ∈ Rn | V (x) ≤ c} is compact for every c > 0.
All solutions starting Ωc will converge to the origin. For any
point p ∈ Rn , choosing c = V (p) ensures that p ∈ Ωc
⇒ The origin is globally asymptotically stable

– p. 7/5
Terminology
V (0) = 0, V (x) ≥ 0 for x 6= 0 Positive semidefinite
V (0) = 0, V (x) > 0 for x 6= 0 Positive definite
V (0) = 0, V (x) ≤ 0 for x 6= 0 Negative semidefinite
V (0) = 0, V (x) < 0 for x 6= 0 Negative definite
kxk → ∞ ⇒ V (x) → ∞ Radially unbounded
Lyapunov’ Theorem: The origin is stable if there is a
continuously differentiable positive definite function V (x) so
that V̇ (x) is negative semidefinite, and it is asymptotically
stable if V̇ (x) is negative definite. It is globally
asymptotically stable if the conditions for asymptotic
stability hold globally and V (x) is radially unbounded

– p. 8/5
A continuously differentiable function V (x) satisfying the
conditions for stability is called a Lyapunov function. The
surface V (x) = c, for some c > 0, is called a Lyapunov
surface or a level surface
c3
c2
V (x) = c 1

c 1 <c 2 <c 3

– p. 9/5
Why do we need the radial unboundedness condition to
show global asymptotic stability?
It ensures that Ωc = {x ∈ Rn | V (x) ≤ c} is bounded for
every c > 0
Without it Ωc might not bounded for large c
Example
x21 2
V (x) = + x 2
1 + x21
x2 c
c
c
c
c
c
hhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
c
c x1
c
c
c
c
c
– p. 10/5
Quadratic Forms
n X
X n
V (x) = xT P x = pij xi xj , P = PT
i=1 j=1

λmin (P )kxk2 ≤ xT P x ≤ λmax (P )kxk2

P ≥ 0 (Positive semidefinite) if and only if λi (P ) ≥ 0 ∀i
P > 0 (Positive definite) if and only if λi (P ) > 0 ∀i
V (x) is positive definite if and only if P is positive definite
V (x) is positive semidefinite if and only if P is positive
semidefinite
P > 0 if and only if all the leading principal minors of P are
positive

– p. 11/5
Linear Systems
ẋ = Ax
V (x) = xT P x, P = PT > 0
def
V̇ (x) = x P ẋ + ẋ P x = x (P A + A P )x = −xT Qx
T T T T

If Q > 0, then A is Hurwitz

Or choose Q > 0 and solve the Lyapunov equation

P A + AT P = −Q

If P > 0, then A is Hurwitz

Matlab: P = lyap(A′ , Q)

– p. 12/5
Theorem A matrix A is Hurwitz if and only if for any
Q = QT > 0 there is P = P T > 0 that satisfies the
Lyapunov equation

P A + AT P = −Q

Moreover, if A is Hurwitz, then P is the unique solution

Idea of the proof: Sufficiency follows from Lyapunov’s

theorem. Necessity is shown by verifying that
Z ∞
P = exp(AT t)Q exp(At) dt
0

is positive definite and satisfies the Lyapunov equation

– p. 13/5
Example:
ẋ = −g(x)
g(0) = 0; xg(x) > 0, ∀ x 6= 0 and x ∈ (−a, a)
Z x
V (x) = g(y) dy
0

∂V
V̇ (x) = [−g(x)] = −g 2 (x) < 0, ∀ x ∈ (−a, a), x 6= 0
∂x
The origin is asymptotically stable

If xg(x) > 0 for all x 6= 0, use

Z x
V (x) = 12 x2 + g(y) dy
0

– p. 14/5
 Z x
V (x) = 12 x2 + g(y) dy
0

is positive definite for all x and radially unbounded since

V (x) ≥ 12 x2

V̇ (x) = −xg(x) − g 2 (x) < 0, ∀ x 6= 0

The origin is globally asymptotically stable

– p. 15/5
Example: Pendulum without friction

ẋ1 = x2
ẋ2 = − a sin x1

V (x) = a(1 − cos x1 ) + 21 x22

V (0) = 0 and V (x) is positive definite over the domain
−2π < x1 < 2π

V̇ (x) = aẋ1 sin x1 + x2 ẋ2 = ax2 sin x1 − ax2 sin x1 = 0

The origin is stable

Since V̇ (x) ≡ 0, the origin is not asymptotically stable

– p. 16/5
Example: Pendulum with friction

ẋ1 = x2
ẋ2 = − a sin x1 − bx2

1
V (x) = a(1 − cos x1 ) + x22
2
V̇ (x) = aẋ1 sin x1 + x2 ẋ2 = − bx22
The origin is stable

V̇ (x) is not negative definite because V̇ (x) = 0 for x2 = 0

irrespective of the value of x1

– p. 17/5
The conditions of Lyapunov’s theorem are only sufficient.
Failure of a Lyapunov function candidate to satisfy the
conditions for stability or asymptotic stability does not mean
that the equilibrium point is not stable or asymptotically
stable. It only means that such stability property cannot be
established by using this Lyapunov function candidate

Try
1 T
V (x) = 2x P x +" a(1 − cos#x"1 ) #
1 p11 p12 x1
= 2 [x1 x2 ] + a(1 − cos x1 )
p12 p22 x2

p11 > 0, p11 p22 − p212 > 0

– p. 18/5
 V̇ (x) = (p11 x1 + p12 x2 + a sin x1 ) x2
+ (p12 x1 + p22 x2 ) (−a sin x1 − bx2 )
= a(1 − p22 )x2 sin x1 − ap12 x1 sin x1
+ (p11 − p12 b) x1 x2 + (p12 − p22 b) x22

p22 = 1, p11 = bp12 ⇒ 0 < p12 < b, Take p12 = b/2

V̇ (x) = − 12 abx1 sin x1 − 21 bx22

D = {x ∈ R2 | |x1 | < π}
V (x) is positive definite and V̇ (x) is negative definite over D
The origin is asymptotically stable

Read about the variable gradient method in the textbook

– p. 19/5
The Invariance Principle

Example: Pendulum with friction

ẋ1 = x2 , ẋ2 = − a sin x1 − bx2

1
V (x) = a(1 − cos x1 ) + x22
2
V̇ (x) = aẋ1 sin x1 + x2 ẋ2 = − bx22
V̇ (x) is not negative definite because V̇ (x) = 0 for x2 = 0
irrespective of the value of x1

However, near the origin, the solution cannot stay

identically in the set {x2 = 0}

– p. 20/5
Definitions: Let x(t) be a solution of ẋ = f (x)

A point p is said to be a positive limit point of x(t) if there is

a sequence {tn }, with limn→∞ tn = ∞, such that
x(tn ) → p as n → ∞

The set of all positive limit points of x(t) is called the

positive limit set of x(t); denoted by L+

An asymptotically stable equilibrium point is the positive

limit set of every solution starting sufficiently near the
equilibrium point

A stable limit cycle is the positive limit set of every solution

starting sufficiently near the limit cycle

– p. 21/5
A set M is an invariant set with respect to ẋ = f (x) if

x(0) ∈ M ⇒ x(t) ∈ M, ∀ t ∈ R

Examples:
Equilibrium points
Limit Cycles
A set M is a positively invariant set with respect to
ẋ = f (x) if

x(0) ∈ M ⇒ x(t) ∈ M, ∀ t ≥ 0

Example: The set Ωc = {V (x) ≤ c} with V̇ (x) ≤ 0 in Ωc

– p. 22/5
Lemma: If a solution x(t) of ẋ = f (x) is bounded and
belongs to D for t ≥ 0, then its positive limit set L+ is a
nonempty, compact, invariant set. Moreover, x(t)
approaches L+ as t → ∞

LaSalle’s theorem: Let f (x) be a locally Lipschitz function

defined over a domain D ⊂ Rn and Ω ⊂ D be a compact
set that is positively invariant with respect to ẋ = f (x). Let
V (x) be a continuously differentiable function defined over
D such that V̇ (x) ≤ 0 in Ω. Let E be the set of all points in
Ω where V̇ (x) = 0, and M be the largest invariant set in E .
Then every solution starting in Ω approaches M as t → ∞

– p. 23/5
Proof:

V̇ (x) ≤ 0 in Ω ⇒ V (x(t)) is a decreasing

V (x) is continuous in Ω ⇒ V (x) ≥ b = min V (x)

x∈Ω

⇒ lim V (x(t)) = a
t→∞

x(t) ∈ Ω ⇒ x(t) is bounded ⇒ L+ exists & L+ ⊂ Ω

x(t) approaches L+ as t → ∞ ⇒ V (x) = a on L+

L+ invariant ⇒ V̇ (x) = 0, ∀ x ∈ L+
L+ ⊂ M ⊂ E ⊂ Ω
x(t) approaches L+ ⇒ x(t) approaches M (as t → ∞)

– p. 24/5
Theorem: Let f (x) be a locally Lipschitz function defined
over a domain D ⊂ Rn ; 0 ∈ D . Let V (x) be a continuously
differentiable positive definite function defined over D such
that V̇ (x) ≤ 0 in D . Let S = {x ∈ D | V̇ (x) = 0}
If no solution can stay identically in S other than
x(t) ≡ 0, then the origin is asymptotically stable

Moreover, if Γ ⊂ D is compact and positively invariant,

then it is a subset of the region of attraction

Furthermore, if D = Rn and V (x) is radially

unbounded, then the origin is globally asymptotically
stable

– p. 25/5
 Example:
ẋ1 = x2
ẋ2 = −h1 (x1 ) − h2 (x2 )
hi (0) = 0, yhi (y) > 0, for 0 < |y| < a
Z x1
V (x) = h1 (y) dy + 12 x22
0

D = {−a < x1 < a, −a < x2 < a}

V̇ (x) = h1 (x1 )x2 +x2 [−h1 (x1 )−h2 (x2 )] = −x2 h2 (x2 ) ≤ 0

V̇ (x) = 0 ⇒ x2 h2 (x2 ) = 0 ⇒ x2 = 0
S = {x ∈ D | x2 = 0}

– p. 26/5
 ẋ1 = x2 , ẋ2 = −h1 (x1 ) − h2 (x2 )

x2 (t) ≡ 0 ⇒ ẋ2 (t) ≡ 0 ⇒ h1 (x1 (t)) ≡ 0 ⇒ x1 (t) ≡ 0

The only solution that can stay identically in S is x(t) ≡ 0

Thus, the origin is asymptotically stable

Ry
Suppose a = ∞ and 0 h1 (z) dz → ∞ as |y| → ∞
R x1
Then, D = R2 and V (x) = 0 h1 (y) dy + 21 x22 is radially
unbounded. S = {x ∈ R2 | x2 = 0} and the only solution
that can stay identically in S is x(t) ≡ 0

The origin is globally asymptotically stable

– p. 27/5
Example: m-link Robot Manipulator

H
Load
H











q

2 



 
 r


 

  q
r 1


Two-link Robot Manipulator

– p. 28/5
 M (q)q̈ + C(q, q̇)q̇ + D q̇ + g(q) = u

q is an m-dimensional vector of joint positions

u is an m-dimensional control (torque) inputs

M = M T > 0 is the inertia matrix

C(q, q̇)q̇ accounts for centrifugal and Coriolis forces

(Ṁ − 2C)T = −(Ṁ − 2C)

D q̇ accounts for viscous damping; D = D T ≥ 0

g(q) accounts for gravity forces

– p. 29/5
Investigate the use of the (PD plus gravity compensation)
control law

u = g(q) − Kp (q − q ∗ ) − Kd q̇

to stabilize the robot at a desired position q ∗ , where Kp and

Kd are symmetric positive definite matrices

e = q − q∗ , ė = q̇

M ë = M q̈
= −C q̇ − D q̇ − g(q) + u
= −C q̇ − D q̇ − Kp (q − q ∗ ) − Kd q̇
= −C ė − D ė − Kp e − Kd ė

– p. 30/5
 M ë = −C ė − D ė − Kp e − Kd ė

V = 12 ėT M (q)ė + 12 eT Kp e

V̇ = ėT M ë + 12 ėT Ṁ ė + eT Kp ė

= −ėT C ė − ėT D ė − ėT Kp e − ėT Kd ė

+ 12 ėT Ṁ ė + eT Kp ė

1 T
= 2 ė (Ṁ − 2C)ė − ėT (Kd + D)ė

= −ėT (Kd + D)ė ≤ 0

– p. 31/5
(Kd + D) is positive definite

V̇ = −ėT (Kd + D)ė = 0 ⇒ ė = 0

M ë = −C ė − D ė − Kp e − Kd ė
ė(t) ≡ 0 ⇒ ë(t) ≡ 0 ⇒ Kp e(t) ≡ 0 ⇒ e(t) ≡ 0
By LaSalle’s theorem the origin (e = 0, ė = 0) is globally
asymptotically stable

– p. 32/5
Region of Attraction

Lemma: If x = 0 is an asymptotically stable equilibrium

point for ẋ = f (x), then its region of attraction RA is an
open, connected, invariant set. Moreover, the boundary of
RA is formed by trajectories

– p. 33/5
Example
ẋ1 = −x2
ẋ2 = x1 + (x21 − 1)x2

4
x
2
2

0
x
1
−2

−4
−4 −2 0 2 4

– p. 34/5
Example
ẋ1 = x2
ẋ2 = −x1 + 13 x31 − x2

4
x
2
2

0
x
1
−2

−4
−4 −2 0 2 4

– p. 35/5
Estimates of the Region of Attraction: Find a subset of the
region of attraction

Warning: Let D be a domain with 0 ∈ D such that for all

x ∈ D , V (x) is positive definite and V̇ (x) is negative
definite (or negative semidefinite and no solution can stay in
{x ∈ D | V̇ (x) = 0} other than x(t) ≡ 0)

Is D a subset of the region of attraction?

NO
Why?

– p. 36/5
Example: Reconsider

ẋ1 = x2
1 3 def
ẋ2 = −x1 + x
3 1
− x2 = −h1 (x1 ) − h2 (x2 )
Z x1
V (x) = (y − 13 y 3 ) + 12 x22 dy = 1 2
2 x1 − 1 4
12 x1 + 12 x22
0

V̇ (x) = (x1 − 13 x31 )x2 + x2 (−x1 + 13 x31 − x2 ) = −x22

In the set √ √
D = {− 3 < x1 < 3}
V (x) is positive definite and x2 (t) ≡ 0 ⇒ x(t) ≡ 0

Is D a subset of the region of attraction?

– p. 37/5
The simplest estimate is the bounded component of
{V (x) < c} where c = minx∈∂D V (x)

Example (Revisited)

ẋ1 = x2
ẋ2 = −x1 + 13 x31 − x2

V (x) = 12 x21 − 12
1 4
x1 + 12 x22
√ √
D = {− 3 < x1 < 3}
3
min V (x) = 4
|x1 |=3

The region of attraction is estimated by {V (x) < 34 }

– p. 38/5
 x’=y
y ’ = − x + x3/3 − y

0
y

−1

−2

−3

−4

−4 −3 −2 −1 0 1 2 3 4
x

– p. 39/5
Finding the estimate {V (x) < c} for a quadratic Lyapunov
function V (x) = xT P x, where P = P T > 0

For D = {kxk < r}, min xT P x = λmin (P )r 2

kxk=r

r2
For D = {|bT x| < r}, min xT P x =
|bT x|=r bT P −1 b

For D = {|bTi x| < ri , i = 1, . . . , p},

ri2
Take c = min ≤ min xT P x
1≤i≤p bTi P −1 bi x∈∂D

– p. 40/5
Example
ẋ1 = −x2
ẋ2 = x1 + (x21 − 1)x2

V (x) = 1.5x21 − x1 x2 + x22

√
5
V̇ (x) = −(x21 + x22 ) − x21 x2 (x1 − 2x2 ) 2
≤ −kxk + kxk4
2
2 def 2
2
V̇ (x) < 0 for 0 < kxk < √ = r
5
2 2
Take c = λmin (P )r = 0.691 × √ = 0.618
5
{V (x) < c} is an estimate of the region of attraction

– p. 41/5
Example

ẋ1 = x2
ẋ2 = −4(x1 + x2 ) − h(x1 + x2 )

h(0) = 0; uh(u) ≥ 0, ∀ |u| ≤ 1

" #
T T 2 1
V (x) = x P x = x x = 2x21 + 2x1 x2 + x22
1 1

V̇ (x) = (4x1 + 2x2 )ẋ1 + 2(x1 + x2 )ẋ2

= −2x21 − 6(x1 + x2 )2 − 2(x1 + x2 )h(x1 + x2 )
≤ −2x21 − 6(x1 + x2 )2 , ∀ |x1 + x2 | ≤ 1

– p. 42/5
V̇ (x) is negative definite in {|x1 + x2 | ≤ 1}
" #
1
b= , {|x1 + x2 | ≤ 1} = {|xT b| ≤ 1}
1

T 1
c= min x Px = =1
|x1 +x2 |=1 bT P −1 b
{V (x) < 1} is an estimate of the region of attraction

Another estimate: According to LaSalle’s theorem any

compact positively invariant subset of {|x1 + x2 | ≤ 1} is a
subset of the region of attraction

– p. 43/5
 σ = x1 + x2

d
σ 2 = 2σx2 − 8σ 2 − 2σh(σ) ≤ 2σx2 − 8σ 2 , ∀ |σ| ≤ 1
dt
d
On σ = 1, σ 2 ≤ 2x2 − 8 ≤ 0, ∀ x2 ≤ 4
dt
d
On σ = −1, σ 2 ≤ −2x2 − 8 ≤ 0, ∀ x2 ≥ −4
dt
c1 = V (x)|x1 =−3,x2 =4 = 10, c2 = V (x)|x1 =3,x2 =−4 = 10

Γ = {V (x) ≤ 10 and |x1 + x2 | ≤ 1}

is a subset of the region of attraction

– p. 44/5
 5 x
2
(−3,4)

V(x) = 10
0 x1

V(x) = 1

(3,−4)
−5
−5 0 5
– p. 45/5
Exponential Stability:

Theorem: Let f (x) be a locally Lipschitz function defined

over a domain D ⊂ Rn ; 0 ∈ D . Let V (x) be a
continuously differentiable function such that

k1 kxka ≤ V (x) ≤ k2 kxka

V̇ (x) ≤ −k3 kxka

for all x ∈ D , where k1 , k2 , k3 , and a are positive
constants. Then, the origin is an exponentially stable
equilibrium point of ẋ = f (x). If the assumptions hold
globally, the origin will be globally exponentially stable

– p. 46/5
Proof: Choose c > 0 small enough that

{k1 kxka ≤ c} ⊂ D

V (x) ≤ c ⇒ k1 kxka ≤ c
Ωc = {V (x) ≤ c} ⊂ { k1 kxka ≤ c} ⊂ D
Ωc is compact and positively invariant; ∀ x(0) ∈ Ωc

a k3
V̇ ≤ −k3 kxk ≤− V
k2
dV k3
≤− dt
V k2
V (x(t)) ≤ V (x(0))e−(k3 /k2 )t

– p. 47/5
  1/a
V (x(t))
kx(t)k ≤
k1
" #1/a
V (x(0))e−(k3 /k2 )t
≤
k1
" #1/a
k2 kx(0)ka e−(k3 /k2 )t
≤
k1
 1/a
k2 −γt k3
= e kx(0)k, γ=
k1 k2 a

– p. 48/5
Example
ẋ1 = x2 , ẋ2 = −h(x1 ) − x2
c1 y 2 ≤ yh(y) ≤ c2 y 2 , ∀ y, c2 > c1 > 0
" # Z x1
1 T 1 1
V (x) = 2 x x +2 h(y) dy
1 2 0
| {z }
1 2 2
2 x 1 +x 1 x2 +x2

Z x1
c1 x21 ≤ 2 h(y) dy ≤ c2 x21
0

V̇ = [x1 + x2 + 2h(x1 )]x2 + [x1 + 2x2 ][−h(x1 ) − x2 ]

= −x1 h(x1 ) − x22 ≤ −c1 x21 − x22
The origin is globally exponentially stable

– p. 49/5
Converse Lyapunov Theorem–Exponential Stability

Let x = 0 be an exponentially stable equilibrium point for

the system ẋ = f (x), where f is continuously differentiable
on D = {kxk < r}. Let k, λ, and r0 be positive constants
with r0 < r/k such that

kx(t)k ≤ kkx(0)ke−λt , ∀ x(0) ∈ D0 , ∀ t ≥ 0

where D0 = {kxk < r0 }. Then, there is a continuously

differentiable function V (x) that satisfies the inequalities

– p. 50/5
 c1 kxk2 ≤ V (x) ≤ c2 kxk2

∂V
f (x) ≤ −c3 kxk2
∂x

∂V
≤ c4 kxk
∂x

for all x ∈ D0 , with positive constants c1 , c2 , c3 , and c4

Moreover, if f is continuously differentiable for all x, globally
Lipschitz, and the origin is globally exponentially stable,
then V (x) is defined and satisfies the aforementioned
inequalities for all x ∈ Rn

– p. 51/5
Idea of the proof: Let ψ(t; x) be the solution of

ẏ = f (y), y(0) = x

Take
Z δ
V (x) = ψ T (t; x) ψ(t; x) dt, δ>0
0

– p. 52/5
Converse Lyapunov Theorem–Asymptotic Stability

Let x = 0 be an asymptotically stable equilibrium point for

ẋ = f (x), where f is locally Lipschitz on a domain
D ⊂ Rn that contains the origin. Let RA ⊂ D be the region
of attraction of x = 0. Then, there is a smooth, positive
definite function V (x) and a continuous, positive definite
function W (x), both defined for all x ∈ RA , such that

V (x) → ∞ as x → ∂RA

∂V
f (x) ≤ −W (x), ∀ x ∈ RA
∂x
and for any c > 0, {V (x) ≤ c} is a compact subset of RA
When RA = Rn , V (x) is radially unbounded

– p. 53/5