Professional Documents
Culture Documents
Introduction To Nonlinear Control Lecture # 4 Passivity
Introduction To Nonlinear Control Lecture # 4 Passivity
Introduction To Nonlinear Control Lecture # 4 Passivity
Lecture # 4
Passivity
– p. 1/6
Memoryless Functions
+ y -
P
y
PP
u P P
P
u
(a)
(b)
power inflow = uy
Resistor is passive if uy ≥ 0
– p. 2/6
y y y
u u u
– p. 3/6
Definition: y = h(t, u) is
passive if uT y ≥ 0
lossless if uT y = 0
– p. 4/6
Sector Nonlinearity: h belongs to the sector [α, β]
(h ∈ [α, β]) if
αu2 ≤ uh(t, u) ≤ βu2
y=βu
y=β u
y y
y=αu
u u
y=αu
[K1 , ∞] if uT [h(t, u) − K1 u] ≥ 0
– p. 6/6
Example
" #
h1 (u1 )
h(u) = , hi ∈ [αi , βi ], βi > αi i = 1, 2
h2 (u2 )
" # " #
α1 0 β1 0
K1 = , K2 =
0 α2 0 β2
h ∈ [K1 , K2 ]
" #
β1 − α1 0
K = K2 − K1 =
0 β2 − α2
– p. 7/6
Example
kh(u) − Luk ≤ γkuk
K1 = L − γI, K2 = L + γI
K = K2 − K1 = 2γI
– p. 8/6
A function in the sector [K1 , K2 ] can be transformed into a
function in the sector [0, ∞] by input feedforward followed
by output feedback
+ +
- - - y = h(t, u) - -
K −1
+6 −6
- K1
Feedforward K −1 Feedback
[K1 , K2 ] [0, K] [0, I] [0, ∞]
−→ −→ −→
– p. 9/6
State Models
v2 = h2(i2) L
y
B B B
i2 + v2 iL
- - BB BB BB -
+ + + +
u v1 i1 = h1(v1) vC C v3 i3 = h3(v3)
P
P P
P
P
P
PP PP
P
P
PP PP
P
P
P P
i1 ?
i3
?
Lẋ1 = u − h2 (x1 ) − x2
C ẋ2 = x1 − h3 (x2 )
y = x1 + h1 (u)
– p. 10/6
V (x) = 12 Lx21 + 12 Cx22
Z t
u(s)y(s) ds ≥ V (x(t)) − V (x(0))
0
– p. 11/6
uy = V̇ + uh1 (u) + x1 h2 (x1 ) + x2 h3 (x2 )
uy ≥ V̇ + uh1 (u)
– p. 12/6
Case 3: If h1 = 0 and h2 ∈ (0, ∞], then
uy ≥ V̇ + x1 h2 (x1 ) + x2 h3 (x2 )
T ∂V
u y ≥ V̇ = f (x, u), ∀ (x, u)
∂x
Moreover, it is said to be
lossless if uT y = V̇
– p. 14/6
output strictly passive if uT y ≥ V̇ + y T ρ(y) for some
function ρ such that y T ρ(y) > 0, ∀ y = 6 0
Example
ẋ = u, y=x
V (x) = 12 x2 ⇒ uy = V̇ ⇒ Lossless
– p. 15/6
Example
ẋ = u, y = x + h(u), h ∈ [0, ∞]
ẋ = −h(x) + u, y = x, h ∈ [0, ∞]
– p. 16/6
Example
ẋ = u, y = h(x), h ∈ [0, ∞]
Z x
V (x) = h(σ) dσ ⇒ V̇ = h(x)ẋ = yu ⇒ Lossless
0
Example
yu = V̇ + xh(x) ⇒ Passive
h ∈ (0, ∞] ⇒ Strictly passive
– p. 17/6
Positive Real Transfer Functions
– p. 18/6
Scalar Case (p = 1):
Re[G(jω)] ≥ 0, ∀ ω ∈ [0, ∞)
– p. 19/6
Lemma: A scalar transfer function G(s) is strictly positive
real if and only if
G(s) is Hurwitz
Re[G(jω)] > 0, ∀ ω ∈ [0, ∞)
G(∞) > 0 or
– p. 20/6
Example:
1
G(s) =
s
has a simple pole at s = 0 whose residue is 1
1
Re[G(jω)] = Re = 0, ∀ ω 6= 0
jω
– p. 21/6
Example:
1
G(s) = , a > 0, is Hurwitz
s+a
a
Re[G(jω)] = > 0, ∀ ω ∈ [0, ∞)
ω2 + a2
ω2a
lim ω 2 Re[G(jω)] = lim = a > 0 ⇒ G is SPR
ω→∞ ω→∞ ω 2 + a2
Example:
1 1 − ω2
G(s) = , Re[G(jω)] =
s2 +s+1 (1 − ω 2 )2 + ω 2
G is not PR
– p. 22/6
Positive Real Lemma: Let
P A + AT P = −LT L
P B = C T − LT W
W T W = D + DT
– p. 23/6
Kalman–Yakubovich–Popov Lemma: Let
P A + AT P = −LT L − εP
P B = C T − LT W
W T W = D + DT
– p. 24/6
Lemma: The linear time-invariant minimal realization
ẋ = Ax + Bu
y = Cx + Du
with
G(s) = C(sI − A)−1 B + D
is
passive if G(s) is positive real
– p. 25/6
T ∂V
u y− (Ax + Bu)
∂x
= uT (Cx + Du) − xT P (Ax + Bu)
= uT Cx + 12 uT (D + D T )u
− 12 xT (P A + AT P )x − xT P Bu
= uT (B T P + W T L)x + 12 uT W T W u
+ 12 xT LT Lx + 12 εxT P x − xT P Bu
= 1
2 (Lx + W u)T (Lx + W u) + 12 εxT P x ≥ 12 εxT P x
– p. 26/6
Connection with Lyapunov Stability
Proof:
T ∂V ∂V
u y≥ f (x, u) ⇒ f (x, 0) ≤ 0
∂x ∂x
– p. 27/6
Lemma: If the system
T ∂V ∂V
u y≥ f (x, u) + ψ(x) ⇒ f (x, 0) ≤ −ψ(x)
∂x ∂x
Why is V (x) positive definite? Let φ(t; x) be the solution
of ż = f (z, 0), z(0) = x
– p. 28/6
V̇ ≤ −ψ(x)
Z τ
V (φ(τ, x)) − V (x) ≤ − ψ(φ(t; x)) dt, ∀ τ ∈ [0, δ]
0
Z τ
V (φ(τ, x)) ≥ 0 ⇒ V (x) ≥ ψ(φ(t; x)) dt
0
Z τ
V (x̄) = 0 ⇒ ψ(φ(t; x̄)) dt = 0, ∀ τ ∈ [0, δ]
0
– p. 29/6
Definition: The system
Linear Systems
ẋ = Ax, y = Cx
– p. 30/6
Lemma: If the system
T ∂V T ∂V
u y≥ f (x, u) + y ρ(y) ⇒ f (x, 0) ≤ −y T ρ(y)
∂x ∂x
– p. 32/6
Feedback Systems
u1- e
n 1-
y1 -
+ H1
−6
y2 e +
? u2
2 n+
H2
yi = hi (t, ei )
– p. 33/6
Passivity Theorems
– p. 34/6
Theorem 6.4: Consider the feedback connection of a
strictly passive dynamical system with a passive
memoryless function. When u = 0, the origin of the
closed-loop system is uniformly asymptotically stable. if the
storage function for the dynamical system is radially
unbounded, the origin will be globally uniformly
asymptotically stable
– p. 35/6
eT1 y1 +eT2 y2 = (u1 −y2 )T y1 +(u2 +y1 )T y2 = uT1 y1 +uT2 y2
V̇ = 0 ⇒ x1 = 0 and y2 = 0
y2 (t) ≡ 0 ⇒ e1 (t) ≡ 0 ( & x1 (t) ≡ 0) ⇒ y1 (t) ≡ 0
y1 (t) ≡ 0 ⇒ e2 (t) ≡ 0
By zero-state observability of H2 : y2 (t) ≡ 0 ⇒ x2 (t) ≡ 0
Apply the invariance principle
– p. 36/6
Example
ẋ1 = x2 ẋ3 = x4
ẋ2 = −ax1 − kx2 + e1 ẋ4 = −bx3 − x34 + e2
3
y1 = x2 + e1 y = x
2 4
| {z } | {z }
H
H
1 2
a, b, k > 0
V1 = 14 ax41 + 12 x22
– p. 37/6
V2 = 21 bx23 + 12 x24
– p. 38/6
Loop Transformation
+ +
- - - y = h(t, u) - -
K −1
+6 −6
- K1
– p. 39/6
+h
- - H1 -
−6
H2
H1 is a dynamical system
H2 is a memoryless function in the sector [K1 , K2 ]
– p. 40/6
+h
- +h -
- -
H1
−6 −6
K1
+
h H2
−
6
K1
– p. 41/6
+h
- +h -
- - K -
H1
−6 −6
K1
+
h H2 K −1
−
6
K1
– p. 42/6
H̃1
+h
- +h -
- - K ?
+
- h -
H1
−6 −6 +
K1
+
+
h H2 K −1
h
−
6 +6
K1
H̃2
– p. 43/6
Example
ẋ1 = x2
ẋ2 = −h(x1 ) + bx2 + e1 y2 = σ(e2 )
| {z }
y1 = x2
H2
| {z }
H1
– p. 44/6
Assume a = α − b > 0 and show that H̃1 is strictly passive
Z x1
V1 = k h(s) ds + xT P x
0
Z x1
V1 = k h(s) ds + p11 x21 + 2p12 x1 x2 + p22 x22
0
– p. 45/6
V̇ = −2p12 x1 h(x1 ) − (ka − 2p12 )x22
+ kx2 ẽ1 + 2p12 x1 ẽ1
= −2p12 x1 h(x1 ) − (ka − 2p12 )x22
+ (kx2 + ẽ1 )ẽ1 − ẽ21 + 2p12 x1 ẽ1
ỹ1 ẽ1 = V̇ + 2p12 x1 h(x1 ) + (ka − 2p12 )x22
+ (ẽ1 − p12 x1 )2 − p212 x21
≥ V̇ + p12 (2α1 − p12 )x21 + (ka − 2p12 )x22
ak k
Take 0 < p12 < min , 2α1 ⇒ p212 < 2p12 = p11 p22
2 2
H̃1 is strictly passive. By Theorem 6.4 the origin is globally
asymptotically stable (when u = 0)
– p. 46/6
Absolute Stability
r + u
- -
y -
G(s)
−6
ψ(·)
– p. 47/6
Circle Criterion
P A + AT P = −LT L − εP
P B = C T − LT W
W T W = D + DT
V (x) = 12 xT P x
– p. 48/6
1 T 1 T
V̇ = 2
x P ẋ + 2
ẋ P x
1 T T T
= 2 x (P A + A P )x + x P Bu
= − 12 xT LT Lx − 12 εxT P x + xT (C T − LT W )u
= − 12 xT LT Lx − 12 εxT P x + (Cx + Du)T u
T T T
− u Du − x L W u
uT Du = 21 uT (D + D T )u = 12 uT W T W u
V̇ = − 12 εxT P x − 1
2
(Lx + W u)T (Lx + W u) − y T ψ(y)
y T ψ(y) ≥ 0 ⇒ V̇ ≤ − 12 εxT P x
The origin is globally exponentially stable
– p. 49/6
What if ψ ∈ [K1 , ∞]?
-
+f - G(s) - -
+f -+f -
G(s) -
− − −
6 6 6
K1
ψ(·) ψ(·)
+
f
−
6
K1 ψ̃(·)
– p. 50/6
What if ψ ∈ [K1 , K2 ]?
-
+f - G(s) - -
+f -+f -
G(s) - K -?+ -
f
− − − +
6 6 6
K1
ψ(·) ψ(·) −1
+
f +
f
− K +
6 6
K1 ψ̃(·)
– p. 51/6
I + KG(s)[I + K1 G(s)]−1 = [I + K2 G(s)][I + K1 G(s)]−1
– p. 52/6
Case 1: α > 0
By the Nyquist criterion
1 + βG(s) 1 βG(s)
= +
1 + αG(s) 1 + αG(s) 1 + αG(s)
– p. 53/6
"1 #
β + G(jω)
Re 1 > 0, ∀ ω ∈ [0, ∞]
α + G(jω)
D(α,β) q
θ θ
2 1
−1/α −1/β
1 + βG(s)
– p. 55/6
Case 3: α < 0 < β
"1 #
1 + βG(jω) β + G(jω)
Re > 0 ⇔ Re 1 <0
1 + αG(jω) α + G(jω)
The Nyquist plot of G(jω) must lie inside the disk D(α, β).
The Nyquist plot cannot encircle the point −(1/α) + j0.
From the Nyquist criterion, G(s) must be Hurwitz
– p. 56/6
Example
4
G(s) =
(s + 1)( 12 s + 1)( 13 s + 1)
6
Im G
4
0
Re G
−2
−4
−5 0 5
– p. 57/6
Apply Case 3 with center (0, 0) and radius = 4
– p. 58/6
Example
4
G(s) =
(s − 1)( 12 s + 1)( 13 s + 1)
0.4 Im G
0.2
G is not Hurwitz
0
Re G
−0.2
Apply Case 1
−0.4
−4 −2 0
Center = (−3.2, 0), Radius = 0.168 ⇒ [0.2969, 0.3298]
– p. 59/6
Popov Criterion
-
+f -
G(s) -
−
6
ψ(·)
ẋ = Ax + Bu
y = Cx
ui = −ψi (yi ), 1 ≤ i ≤ p
ψi ∈ [0, ki ], 1 ≤ i ≤ p, (0 < ki ≤ ∞)
G(s) = C(sI − A)−1 B
Γ = diag(γ1 , . . . , γp ), M = diag(1/k1 , · · · , 1/kp )
– p. 60/6
- M
H̃1
- ?
+ +
- i - G(s) - (I + sΓ) i -
−6 +
+
ψ(·) (I + sΓ)−1
i
+6
H̃2
M
P A + AT P = −LT L − εP
P B = (C + ΓCA)T − LT W
W T W = 2M + ΓCB + B T C T Γ
– p. 62/6
H̃2 consists of p decoupled components:
1
γi żi = −zi + ψi (zi ) + ẽ2i , ỹ2i = ψi (zi )
ki
Z zi
V2i = γi ψi (σ) dσ
0
h i
1
V̇2i = γi ψi (zi )żi = ψi (zi ) −zi + ki ψi (zi ) + ẽ2i
1
= y2i e2i + ki ψi (zi ) [ψi (zi ) − ki z i ]
– p. 63/6
p
X Z yi
Use V = 21 xT P x + γi ψi (σ) dσ
i=1 0
1 T 1 T T (y)Γẏ
V̇ = 2
x P ẋ + 2
ẋ P x + ψ
1 T T P )x + xT P Bu
= 2
x (P A + A
+ ψ T (y)ΓC(Ax + Bu)
= − 12 xT LT Lx − 12 εxT P x
+ xT (C T + AT C T Γ − LT W )u
+ ψ T (y)ΓCAx + ψ T (y)ΓCBu
– p. 64/6
V̇ = − 21 εxT P x − 12 (Lx + W u)T (Lx + W u)
− ψ(y)T [y − M ψ(y)]
≤ − 12 εxT P x − ψ(y)T [y − M ψ(y)]
– p. 65/6
Scalar case
1
+ (1 + sγ)G(s)
k
is SPR if G(s) is Hurwitz and
1
+ Re[G(jω)] − γωIm[G(jω)] > 0, ∀ ω ∈ [0, ∞)
k
If
1
lim + Re[G(jω)] − γωIm[G(jω)] =0
ω→∞ k
we also need
2 1
lim ω + Re[G(jω)] − γωIm[G(jω)] > 0
ω→∞ k
– p. 66/6
1
+ Re[G(jω)] − γωIm[G(jω)] > 0, ∀ ω ∈ [0, ∞)
k
−1/k Re[G(jω)]
Popov Plot
– p. 67/6
Example
– p. 68/6
The system is absolutely stable for ψ ∈ [0, ∞] (h ∈ [α, ∞])
ω Im G
0.2 slope=1
0
Re G
−0.2
−0.4
−0.6
−0.8
−1
−0.5 0 0.5 1
1 α − ω2 √
+ > 0, ∀ ω ∈ [0, ∞], for k < 1 + 2 α
k (α − ω 2 )2 + ω 2
– p. 69/6