Lecture 5 — Input–output stability

or
“How to make a circle out of the point −1 + 0i, and different ways to
stay away from it ...”

k2 y f (y)
r y k1 y
G(s) − k11 − k12
− y

f (·)
G(iω)

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Today’s Goal

To understand

signal norms
system gain
bounded input bounded output (BIBO) stability

To be able to analyze stability using

the Small Gain Theorem,

the Circle Criterion,
Passivity

Material

[Glad & Ljung]: Ch 1.5-1.6, 12.3 [Khalil]: Ch 5–7.1

lecture slides

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 History

f (y)
r y
G(s)
− y

f (·)

For what G(s) and f (·) is the closed-loop system stable?

Lur’e and Postnikov’s problem (1944)

Aizerman’s conjecture (1949) (False!)
Kalman’s conjecture (1957) (False!)
Solution by Popov (1960) (Led to the Circle Criterion)

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Gain

Idea: Generalize static gain to nonlinear dynamical systems

u y
S

The gain γ of S measures the largest amplification from u to y

Here S can be a constant, a matrix, a linear time-invariant system, a
nonlinear system, etc
Question: How should we measure the size of u and y ?

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Norms

A norm k · k measures size.

A norm is a function from a space Ω to R+ , such that for all x, y ∈ Ω

kxk ≥ 0 and kxk = 0 ⇔ x = 0

kx + yk ≤ kxk + kyk
kαxk = |α| · kxk, for all α ∈ R

Examples q
Euclidean norm: kxk = x21 + · · · + x2n
Max norm: kxk = max{|x1 |, . . . , |xn |}

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Signal Norms

A signal x(t) is a function from R+ to Rd .

A signal norm is a way to measure the size of x(t).

Examples qR
∞
2-norm (energy norm): kxk2 = 0 |x(t)|2 dt
sup-norm: kxk∞ = supt∈R+ |x(t)|

The space of signals with kxk2 < ∞ is denoted L2 .

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Parseval’s Theorem

Theorem If x, y ∈ L2 have the Fourier transforms

Z ∞ Z ∞
X(iω) = e−iωt x(t)dt, Y (iω) = e−iωt y(t)dt,
0 0

then Z ∞ Z ∞
1
T
y (t)x(t)dt = Y ∗ (iω)X(iω)dω.
0 2π −∞
In particular
Z ∞ Z ∞
1
kxk22 = 2
|x(t)| dt = |X(iω)|2 dω.
0 2π −∞

kxk2 < ∞ corresponds to bounded energy.

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 System Gain

A system S is a map between two signal spaces: y = S(u).

u y
S

kyk2 kS(u)k2
The gain of S is defined as γ(S) = sup = sup
u∈L2 kuk2 u∈L2 kuk2

Example The gain of a static relation y(t) = αu(t) is

kαuk2 |α|kuk2
γ(α) = sup = sup = |α|
u∈L2 kuk2 u∈L2 kuk2

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Example—Gain of a Stable Linear System

1
10

γ(G) |G(iω)|
0
10

kGuk2
γ G = sup = sup |G(iω)| −1

u∈L2 kuk2 ω∈(0,∞)

10

−2
10 −1 0 1
10 10 10

Proof: Assume |G(iω)| ≤ K for ω ∈ (0, ∞). Parseval’s theorem gives

Z
1 ∞
kyk22 = |Y (iω)|2 dω
2π −∞
Z
1 ∞
= |G(iω)|2 |U (iω)|2 dω ≤ K 2 kuk22
2π −∞

This proves that γ(G) ≤ K . See [Khalil, Appendix C.10] for a proof of
the equality.

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

2 minute exercise: Show that γ(S1 S2 ) ≤ γ(S1 )γ(S2 ).

u y
S2 S1

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Example—Gain of a Static Nonlinearity

|f (x)| ≤ K|x|, f (x∗ ) = Kx∗

Kx
f (x)
u(t) y(t)
f (·) x
x∗

Z ∞ Z ∞


kyk22 = f 2 u(t) dt ≤ K 2 u2 (t)dt = K 2 kuk22
0 0
(
x∗ 0 ≤ t ≤ 1
for u(t) = one has ||y||2 = ||Ku||2 = K||u||2
0 t>1
kyk2
=⇒ γ(f ) = sup = K.
u∈L2 kuk2
Lecture 5 Nonlinear Control and Servo systems (FRTN05)
 BIBO Stability

u y kyk2
S γ(S) = sup
u∈L2 kuk2

Definition
S is bounded-input bounded-output (BIBO) stable if γ(S) < ∞.

Example: If ẋ = Ax is asymptotically stable then

G(s) = C(sI − A)−1 B + D is BIBO stable.

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 The Small Gain Theorem

r1 e1
S1

e2 r2
S2

Theorem
Assume S1 and S2 are BIBO stable. If

γ(S1 )γ(S2 ) < 1

then the closed-loop map from (r1 , r2 ) to (e1 , e2 ) is BIBO stable.

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Proof

qR
T
Define kykT = 0 |y(t)|2 dt. Then kS(y)kT ≤ kSk · kykT .

e1 = r1 + S2 (r2 + S1 (e1 ))
 
ke1 kT ≤ kr1 kT + kS2 k kr2 kT + kS1 k · ke1 kT
kr1 kT + kS2 k · kr2 kT
ke1 kT ≤
1 − kS1 k · kS2 k

This shows bounded gain from (r1 , r2 ) to e1 .

The gain to e2 is bounded in the same way.

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Linear System with Static Nonlinear Feedback (1)

Ky
r y f (y)
G(s)
−
y

f (·)

2 f (y)
G(s) = and 0≤ ≤K
(s + 1)2 y

γ(G) = 2 and γ(f ) ≤ K .

The small gain theorem gives that K ∈ [0, 1/2) implies BIBO stability.

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Example

r y
G(s)
−

1
4 sin(·)

(
ẋ = −x + r − sin(y)/4 2
G(s) =
ẏ = −y + 2x (s + 1)2

The closed loop system is stable by the small gain theorem.

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 The Nyquist Theorem

Ω 1.5

G(s)
1

−
0.5

−0.5
G(Ω)
−1

−1.5

−2

−2 −1.5 −1 −0.5 0 0.5 1

Theorem
If G(s) is stable, then the closed loop system [1 + G(s)]−1 is stable if
and only if the Nyquist curve does not encircle −1
The difference between the number of unstable poles in [1 + G(s)]−1
and the number of unstable poles in G(s) is equal to the number of
times the point −1 is encircled by the Nyquist plot in the clockwise
direction.

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 The Small Gain Theorem can be Conservative

Let f (y) = Ky for the previous system.

0.5

−0.5
G(iω)
−1

−1 −0.5 0 0.5 1 1.5 2

The Nyquist Theorem proves stability when K ∈ [0, ∞).

The Small Gain Theorem proves stability when K ∈ [0, 1/2).

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 The Circle Criterion
Case 1: 0 < k1 ≤ k2 < ∞
k2 y f (y)
r y
replacements k1 y
G(s) − k11 − k12
− y
f (·)
G(iω)

Theorem Consider a feedback loop with y = Gu and

u = −f (y) + r . Assume G(s) is stable and that

f (y)
0 < k1 ≤ ≤ k2 .
y

If the Nyquist curve of G(s) does not intersect or encircle the circle
defined by the points −1/k1 and −1/k2 , then the closed-loop system
is BIBO stable from r to y .

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Other cases
G: stable system

0 < k1 < k2 : Stay outside circle

0 = k1 < k2 : Stay to the right of the line Re s = −1/k2
k1 < 0 < k2 : Stay inside the circle

Other cases: Multiply f and G with −1.

G: Unstable system
To be able to guarantee stability, k1 and k2 must have same sign
(otherwise unstable for k = 0)

0 < k1 < k2 : Encircle the circle p times counter-clockwise (if ω

increasing)
k1 < k2 < 0: Encircle the circle p times counter-clockwise (if ω
increasing)

where p=number of open loop unstable poles

Lecture 5 Nonlinear Control and Servo systems (FRTN05)
 Linear System with Static Nonlinear Feedback (2)

Ky
f (y) 1

0.5
1
y −
0
K
−0.5 G(iω)
−1

−1 −0.5 0 0.5 1 1.5 2

The “circle” is defined by −1/k1 = −∞ and −1/k2 = −1/K .

min Re G(iω) = −1/4

so the Circle Criterion gives that if K ∈ [0, 4) the system is BIBO

stable.

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Example

r y 1
G(s)
− 0.5
1
−
0
K
−0.5 G(iω)
sat(·)
−1

−1 −0.5 0 0.5 1 1.5 2

(
ẋ = −x + r − sat(y) 2
G(s) =
ẏ = −y + 2x (s + 1)2
sat(x)
0≤ ≤1<K
x
The closed loop system is BIBO stable by the circle criterion.

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Proof of the Circle Criterion

Let k = (k1 + k2 )/2 and fe(y) = f (y) − ky . Then

fe(y) k2 − k1
≤ =: R
y 2

e
G −k
r1 e1 y1
G(s)
re1 y1
G
y2 e2 r2
−f (·)
r2
−fe(·)

re1 = r1 − kr2

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Proof of the Circle Criterion (cont’d)

re1 e
G(s)
−k
r2 R
−fe(·)
1
G(iω)

e e= G
SGT gives stability for |G(iω)|R < 1 with G .
1 + kG
1 1
R< = +k
e
|G(iω)| G(iω)

Transform this expression through z → 1/z .

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Lyapunov revisited

Original idea: “Energy is decreasing”

ẋ = f (x), x(0) = x0
V (x(T )) − V (x(0)) ≤ 0
(+some other conditions on V )

New idea: “Increase in stored energy ≤ added energy”

ẋ = f (x, u), x(0) = x0

y = h(x)
Z T
V (x(T )) − V (x(0)) ≤ ϕ(y, u) dt (1)
0 | {z }
external power

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Motivation

Will assume the external power has the form φ(y, u) = y ⊤ u.

Only interested in BIBO behavior. Note that

∃V ≥ 0 with V (x(0)) = 0 and (1)

⇐⇒
Z T
y ⊤ u dt ≥ 0
0

Motivated by this we make the following definition

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Passive System

u y
S

Definition The system S is passive from u to y if

Z T
y ⊤ u dt ≥ 0, for all u and all T > 0
0

and strictly passive from u to y if there ∃ǫ > 0 such that

Z T Z T
y ⊤ u dt ≥ ǫ (|y|2 + |u|2 )dt, for all u and all T > 0
0 0

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 A Useful Notation

Define the scalar product

u y
Z T S
hy, uiT = y(t)⊤ u(t) dt
0

Cauchy-Schwarz inequality:

hy, ui2T ≤ hy, yiT · hu, uiT .

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 2 minute exercise

Assume S1 and S2 are passive. Are then parallel connection and

series connection passive? How about inversion; S1−1 ?

S1
u y u y
S1 S2
S2

u y
S1−1

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Feedback of Passive Systems is Passive
r1 e1 y1
S1
−

y2 e2 r2
S2

If S1 and S2 are passive, then the closed-loop system from (r1 , r2 ) to

(y1 , y2 ) is also passive.

Proof: hy, riT = hy1 , r1 iT + hy2 , r2 iT

= hy1 , r1 − y2 iT + hy2 , r2 + y1 iT
= hy1 , e1 iT + hy2 , e2 iT ≥ 0
Hence, hy, riT ≥ 0 if hy1 , e1 iT ≥ 0 and hy2 , e2 iT ≥ 0

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Passivity of Linear Systems

Theorem An asymptotically stable linear system G(s) is passive if

and only if
Re G(iω) ≥ 0, ∀ω > 0
It is strictly passive if and only if there exists ǫ > 0 such that

Re G(iω) ≥ ǫ(1 + |G(iω)|2 ), ∀ω > 0

Example 0.6

s+1 0.4
G(s) = is passive and
s+2 0.2

strictly passive, 0

1 −0.2
G(iω)
G(s) = is passive but not strictly −0.4
s
passive. 0 0.2 0.4 0.6 0.8 1

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 A Strictly Passive System Has Finite Gain

u y
S

If S is strictly passive, then γ(S) < ∞.

Proof: Note that

q
ǫhy, yiT + ǫhu, uiT ≤ hy, uiT ≤ hy, yiT · hu, uiT ≤ kyk2 · kuk2

Hence, ǫhy, yiT ≤ kyk2 · kuk2 , so letting T → ∞ gives

1
kyk2 ≤ kuk2
ǫ

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 The Passivity Theorem

r1 e1 y1
S1
−

y2 e2 r2
S2

Theorem If S1 is strictly passive and S2 is passive, then the

closed-loop system is BIBO stable from r to y .

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Proof of the Passivity Theorem

S1 strictly passive and S2 passive give

ǫhy1 , y1 iT + ǫhe1 , e1 iT ≤ hy1 , e1 iT + hy2 , e2 iT = hy, riT

Therefore
1
hy1 , y1 iT + hr1 − y2 , r1 − y2 iT ≤ hy, riT
ǫ
or
1
hy, yiT − 2hy2 , r2 iT + hr1 , r1 iT ≤ hy, riT
ǫ
Finally
 
1 1 q
hy, yiT ≤ 2hy2 , r2 iT + hy, riT ≤ 2 + hy, yiT hr, riT
ǫ ǫ

Letting T → ∞ gives kyk2 ≤ Ckrk2 and the result follows

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

Passivity Theorem is a “Small Phase Theorem”

r1 e1 y1
S1
−
y2 e2 r2
S2
φ1 φ2

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Example—Gain Adaptation
Applications in channel estimation in telecommunication, noise
cancelling etc.
replacements
Process
u y
θ∗ G(s)

ym
θ(t) G(s)

Model

Adaptation law:

dθ
= −γu(t)[ym (t) − y(t)], γ > 0.
dt

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

Gain Adaptation—Closed-Loop System

u y
θ∗ G(s)
− γ θ
−
ym s
θ(t) G(s)

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Gain Adaptation is BIBO Stable

(θ − θ ∗ )u ym − y
G(s)

θ∗
− θ γ
−
s

u S

S is passive (Exercise 4.12), so the closed-loop system is BIBO stable

if G(s) is strictly passive.

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Simulation of Gain Adaptation

1
Let G(s) = + ǫ, γ = 1, u = sin t, θ(0) = 0 and γ ∗ = 1
s+1

y , ym
0

−2
0 5 10 15 20

1.5

1 θ

0.5

0
0 5 10 15 20

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Storage Function

Consider the nonlinear control system

ẋ = f (x, u), y = h(x)

A storage function is a C 1 function V : Rn → R such that

V (0) = 0 and V (x) ≥ 0, ∀x 6= 0

V̇ (x) ≤ uT y , ∀x, u

Remark:

V (T ) represents the stored energy in the system

Z T
V (x(T )) ≤ y(t)u(t)dt + V (x(0)) , ∀T > 0
| {z } | {z }
stored energy at t = T
|0 {z } stored energy at t = 0
absorbed energy

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Storage Function and Passivity

Lemma: If there exists a storage function V for a system

ẋ = f (x, u), y = h(x)

with x(0) = 0, then the system is passive.

Proof: For all T > 0,

Z T
hy, uiT = y(t)u(t)dt ≥ V (x(T )) − V (x(0)) = V (x(T )) ≥ 0
0

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Lyapunov vs. Passivity

Storage function is a generalization of Lyapunov function

Lyapunov idea: “Energy is decreasing”

V̇ ≤ 0

Passivity idea: “Increase in stored energy ≤ Added energy”

V̇ ≤ uT y

Lecture 5 Nonlinear Control and Servo systems (FRTN05)

 Example KYP Lemma

Consider an asymptotically stable linear system

ẋ = Ax + Bu, y = Cx

Assume there exists positive definite symmetric matrices P , Q such

that
AT P + P A = −Q, and B T P = C
Consider V = 0.5xT P x. Then

V̇ = 0.5(ẋT P x + xT P ẋ) = 0.5xT (AT P + P A)x + uT B T P x

(2)
= −0.5xT Qx + uT y < uT y, x 6= 0

and hence the system is strictly passive. This fact is part of the
Kalman-Yakubovich-Popov lemma.

Lecture 5 Nonlinear Control and Servo systems (FRTN05)