Control Systems - II Session: 2024-2025

Control Systems - II (Part II)

Control Systems - II Session: 2024-2025
Controllability and Observability

ẋ1 = −x1 + u
ẋ2 = −2x2
ẋ3 = −3x3 + u
ẋ4 = −4x4

y = x1 + x2

x1 → Affected by the Input, Visible in the Output

x2 → Unaffected by the Input, Visible in the Output
x3 → Affected by the Input, Invisible in the Output
x4 → Unaffected by the Input, Invisible in the Output

Figure: Controllability and Observability

Control Systems - II Session: 2024-2025
Controllability and Observability

Controllability and Observability are distinct attributes to

State-Space methods.
It is a way of explaining why a method of designing Compensators
for Unstable Systems by cancelling Unstable poles (poles in the
right half s-plane) by zeros in the right half s-plane is doomed to
fail even if the cancellation is perfect.
A perfect pole-zero cancellation would result in an Unstable System
with a Stable Transfer Function.
The Transfer Function is of lower order than the System.
The Unstable modes are either not capable of being affected by the
input i.e. Uncontrollable or not visible in the output i.e.
Unobservable.
Control Systems - II Session: 2024-2025
Controllability and Observability

Controllability
A System is said to be Controllable if and only if it is possible, by means
of the input, to transfer the System from any initial State x(t) = xt to
any other State xT = x(T ) in a finite time T − t ≥ 0.

Controllability Theorem
A System is Controllable if and only if the matrix,
Z T
P(T , t) = φ(T , λ)B(λ)B 0 (λ)φ0 (T , λ)dλ (1)
t

is non-singular for some T > t where φ(T , t) is the State-Transition

matrix of the System.

The above matrix integral is called the Controllability Grammian.

Control Systems - II Session: 2024-2025
Controllability and Observability

For a Time-Invariant System, the Controllability Grammian is,

Z T Z T −t
A(T −λ) 0 A0 (T −λ) 0
P(T − t) = e BB e dλ = e At BB 0 e A t dt (2)
t 0

or simply,
Z T
0
P(T ) = e At BB 0 e A t dt (3)
0
Sometimes, the matrix used for the Controllability test is expressed as,
Z T
0
P̄(T ) = e −At BB 0 e −A t dt (4)
0

This matrix has the same rank as P(T ).

Just as the output y is not considered in the definition of Controllability,
the input u is generally not considered in defining Observability.
Control Systems - II Session: 2024-2025
Controllability and Observability

Consider the Unforced System,

ẋ(t) = A(t)x(t)
y (t) = C (t)x(t)

Observability
An Unforced System is said to be Observable if and only if it is possible
to determine any arbitrary initial State x(t) = xt bu using only a finite
information of output, y (τ ) for t ≤ τ ≤ T .

Observability Theorem
A System is Observable if and only if the matrix,
Z T
M(T , t) = φ0 (λ, t)C 0 (λ)C (λ)φ(λ, t)dλ (5)
t

is non-singular for some T > t where, φ(T , t) is the State-Transition

Matrix of the System.
The above matrix integral is called the Observability Grammian.
Control Systems - II Session: 2024-2025
Controllability and Observability

For Time-Invariant Systems, the Observability Grammian is,

Z T Z T −t
0 0
M(T − t) = e A (λ−t) C 0 Ce A(λ−t) dλ = e A τ C 0 Ce Aτ dτ (6)
t 0

or simply,
Z T
0
M(T ) = e A τ C 0 Ce Aτ dτ (7)
0
In other form, Z T
0
M̄(T ) = e −A t C 0 Ce −At dt (8)
0

This matrix has the same rank as M(T ).

It is not necessary to evaluate the Controllability Grammian and
Observability Grammian to test for Controllability and Observability.
There are simple algebraic tests for those.
Control Systems - II Session: 2024-2025
Controllability and Observability

Algebraic Controllability Theorem

The Time-Invariant System,

ẋ = Ax + Bu

is Controllable if and only if the rank r (Qc ) of the Controllability Matrix,

Qc = [B AB · · · Ak−1 B] (9)

is equal to the Order, n of the System.

Control Systems - II Session: 2024-2025
Controllability and Observability

Algebraic Observability Theorem

The Unforced Time-Invariant System,

ẋ = Ax, y = Cx

is Observable if and only if the rank r (Qo ) of the Observability Matrix,

Qo = [C > A> C > · · · (A> )k−1 C > ] (10)

is equal to the Order, n of the System.

Control Systems - II Session: 2024-2025
Controllability and Observability

Example

   
0 1 0 0
ẋ =  0 0 1  x + 0 u
−a0 −a1 −a2 1
 
y= 1 0 0 x
       
0 1 0 0 0 1
A= 0 0 1  , B = 0 , AB =  1  , B =  −a2 
−a0 −a1 −a2 1 −a2 a22 − a1
 
0 0 1
Qc = B AB A2 B = 0
 
1 −a2 
0 −a2 a22 − a1

|Qc | = −1 6= 0
The System is Controllable.
Control Systems - II Session: 2024-2025
Controllability and Observability

Example

   
0 1 0 0
ẋ = 0 0 1  x + 0 u
0 −2 −3 1
 
y= 3 4 1 x
       
0 0 0 3 0 0
A> = 1 0 −2 , C > = 4 , A> C > = 1 , (A> )2 C > = −2
0 1 −3 1 1 −2
 
3 0 0
Qo = C > A> C > (A> )2 C > = 4 1 −2
 
1 1 −2

|Qo | = 0
The System is Unobservable.
Control Systems - II Session: 2024-2025
State Feedback Control Design

State Feedback Control

ẋ = Ax + Bu
(11)
y = Cx

Control Law,
u = −Kx (12)
Closed-Loop System Dynamics,

ẋ = Ax + B(−Kx)
= (A − BK )x (13)
y = Cx
Control Systems - II Session: 2024-2025
State Feedback Control Design

Assuming phase-variable form,

   
0 1 0 ··· 0 0
 0
 0 1 ··· 0 

0
 
 · · · · · 
 ; B = · (14)
 

 · · · · ·  ·
−a0 −a1 −a2 · · · −an−1 1
 
C = c1 c2 · · · cn (15)
Characteristic Equation,

s n + an−1 s n−1 + · · · + a1 s + a0 = 0 (16)

Control Systems - II Session: 2024-2025
State Feedback Control Design

Control Law,
u = −Kx (17)
where,  
K = k1 k2 · · · kn

Closed-Loop System Matrix, A − BK =

 
0 1 0 · · · 0

 0 0 1 · · · 0 

 · · · · · 
−(a0 + k1 ) −(a1 + k2 ) −(a2 + k3 ) · −(an−1 + kn )
(18)
Control Systems - II Session: 2024-2025
State Feedback Control Design

Closed-Loop Characteristic Equation,

det(sI − (A − BK ))

= s n +(an−1 +kn )s n−1 +(an−2 +kn−1 )s n−2 +···+(a1 +k2 )s+(a0 +k1 ) = 0
(19)
Desired Characteristic Equation,

s n + dn−1 s n−1 + dn−2 s n−2 + · · · + d2 s 2 + d1 s + d0 = 0 (20)

Feedback Gains,

ki+1 = di − ai , i = 0, 1, 2, ....., n − 1 (21)

Control Systems - II Session: 2024-2025
State Feedback Control Design

Transfer Function,
X (s) 20(s + 5)
= (22)
U(s) s(s + 1)(s + 4)
In state-space representation with phase-variable form,
ẋ1 = x2
ẋ2 = x3
(23)
ẋ3 = −k1 x1 − (k2 + 4)x2 − (k3 + 5)x3 + u
y = 100x1 + 20x2
      
ẋ1 0 1 0 x1 0
ẋ2  =  0 0 1  x2  + 0 u (24)
ẋ3 −k1 −(k2 + 4) −(k3 + 5 x3 1
 
  x1
y = 100 20 0 x2  (25)
x3
Control Systems - II Session: 2024-2025
State Feedback Control Design

 
0 1 0
A − BK =  0 0 1  (26)
−k1 −(k2 + 4) −(k3 + 5)
det(sI − (A − BK )) = s 3 + (k3 + 5)s 2 + (k2 + 4)s + k1 (27)

Required Overshoot = 9.5 %, Required Settling Time = 0.74 sec.

Closed-loop poles → −5.4 ± j7.2, Third pole = -5.1
Desired characteristic equation,
s 3 + 15.9s 2 + 136.08s + 413.1 = 0 (28)
Comparing (54) and (55),

k1 = 413.1, k2 = 132.08, k3 = 10.9

Control Law,
u = −413.1x1 − 132.08x2 − 10.9x3 (29)
Control Systems - II Session: 2024-2025
State Feedback Control Design by Ackermann’s Formula

For Single-Input Single-Output Systems, Ackermann’s Formula may be

used to find the gain K of State Feedback Control.

Desired Characteristic Equation,

p(λ) = λn + αn−1 λn−1 + · · · + α0

State Feedback Gain,

1 Qc−1 p(A)
 
K= 0 0 0 0 ··· (30)

where,
p(A) = An + αn−1 An−1 + · · · + α1 A + α0 I
Control Systems - II Session: 2024-2025
State Feedback Control Design by Ackermann’s Formula

Transfer Function,
Y (s) 1
G (s) = = 2
U(s) s
Desired Characteristic Equation,

p(λ) = λ2 + 2λ + 2

Defining the States, x1 = y , x2 = ẏ

State-Space Model,    
0 1 0
ẋ = x+ u
0 0 1
Controllability Matrix,
 
  0 1
Qc = B AB =
1 0
 
0 1
Qc−1 =
1 0
Control Systems - II Session: 2024-2025
State Feedback Control Design by Ackermann’s Formula

 2      
0 1 0 1 1 0 2 2
p(A) = +2 +2 =
0 0 0 0 0 1 0 2

State Feedback Gain,

  
  0 1 2 2  
K= 0 1 = 2 2
1 0 0 2
Control Systems - II Session: 2024-2025
Observer Design

Figure: State-Space Representation

ẋ = Ax + Bu
(31)
y = Cx
Control Systems - II Session: 2024-2025
Observer Design

Figure: State Feedback Control

ẋ = Ax + Bu
(32)
y = Cx
State Feedback Control Law,

u = −Kx (33)
Control Systems - II Session: 2024-2025
Observer Design

Significance of Observer Design

Control design relies on the information of states for feedback.

Inaccessibility of all states.
Costly measurement.
States can be estimated by designing observer.
Control Systems - II Session: 2024-2025
Observer Design

Observer dynamics,

x̂˙ = Ax̂ + Bu
(34)
y = C x̂

Subtracting (34) from (32),

ẋ − x̂˙ = A(x − x̂)

(35)
y − ŷ = C (x − x̂)
Figure: Open-Loop Observer

The error dynamics (35) is unforced.

Error converges to zero if the plant is stable.
Rate of convergence is the same as that of plant.
Error should converge faster than the closed-loop system
response.
Control Systems - II Session: 2024-2025
Observer Design

The response becomes

faster with feedback.
With feedback, the desired
transient response can be
achieved which is much
faster than that of the plant
or the closed-loop system.
Observer canonical form
gives easy solution for gains.
Observer design is separate Figure: Closed-Loop Observer
from controller design
(Separation Principle).
Control Systems - II Session: 2024-2025
Observer Design

Figure: Closed-Loop Observer

Observer dynamics,
x̂˙ = Ax̂ + Bu + L(y − ŷ )
(36)
ŷ = C x̂
Subtracting (36) from (32),
ẋ − x̂˙ = A(x − x̂) + L(y − ŷ )
(37)
y − ŷ = C (x − x̂)
Control Systems - II Session: 2024-2025
Observer Design

ẋ − x̂˙ = (A − LC )(x − x̂)

(38)
y − ŷ = C (x − x̂)
Defining error variable as,

e = (x − x̂) (39)

Error dynamics,
ė = (A − LC )e
(40)
y − ŷ = Ce
In the above unforced system, the error, e converges to zero if its
eigenvalues are all negative.
Characteristic Equation,

det[λI − (A − LC )] = 0 (41)
Control Systems - II Session: 2024-2025
Observer Design

Assuming observer canonical form,

   
−an−1 1 0 ··· 0 b1
−an−2 0 1 · · · 0  b2 
  

A=  · · · · · ; B =  ·  (42)
 

 · · · · · ·
−a0 0 0 ··· 0 bn
 
C = 1 0 ··· 0 (43)
 
  l1
−(an−1 + l1 ) 1 0 · · · 0 l2 
−(an−2 + l2 ) 0 1 · · · 0  
  ·
A − LC = 
 · · · · · ; L = 

·
 (44)
 · · · · ·  
·
−(a0 + ln ) 0 0 · · · 0
ln
Control Systems - II Session: 2024-2025
Observer Design

Characteristic Equation,

s n + (an−1 + l1 )s n−1 + · · · + (a1 + ln−1 )s + (a0 + ln ) = 0 (45)

Desired Characteristic Equation,

s n + dn−1 s n−1 + dn−2 s n−2 + · · · + d2 s 2 + d1 s + d0 = 0 (46)

Observer Gains,

li = dn−i − an−i , i = 0, 1, 2, ....., n − 1 (47)

Control Systems - II Session: 2024-2025
Observer Design

Transfer Function,
s +4
G (s) = (48)
s3 + 8s 2+ 17s + 10
In state-space representation (observer canonical form),
ẋ1 = −8x1 + x2
ẋ2 = −17x1 + x3 + u
(49)
ẋ3 = −10x1 + 4u
y = x1
      
ẋ1 −8 1 0 x1 0
ẋ2  = −17 0 1 x2  + 1 u (50)
ẋ3 −10 0 0 x3 4
 
  x1
y = 1 0 0 x2  (51)
x3
Control Systems - II Session: 2024-2025
Observer Design

 
−(8 + l1 ) 1 0
A − LC = −(17 + l2 ) 0 1 (52)
−(10 + l3 ) 0 0
det(sI − (A − LC )) = s 3 + (8 + l1 )s 2 + (17 + l2 )s + (10 + l3 ) (53)
Desired characteristic equation of closed-loop system,
(s + 4)(s 2 + 2s + 5) = s 3 + 6s 2 + 13s + 20 = 0 (54)
Dominant poles = −1 ± j2

Observer response is desired to be 10 times faster.

Observer poles = −10 ± j20, Third pole = - 100
Desired characteristic equation,

(s + 100)(s 2 + 20s + 500) = s 3 + 120s 2 + 2500s + 50000 = 0 (55)

Comparing (54) and (55),

l1 = 112, l2 = 2483, l3 = 49990

Control Systems - II Session: 2024-2025
Observer Design by Ackermann’s Formula

For Single-Input Single-Output Systems, Ackermann’s Formula may be

used to find the gain L of Observer.

Desired Characteristic Equation,

p(λ) = λn + βn−1 λn−1 + · · · + β0

State Feedback Gain,

>
L = p(A)Qo−1 0

0 0 0 ··· 1 (56)

where,
p(A) = An + βn−1 An−1 + · · · + β1 A + β0 I
Control Systems - II Session: 2024-2025
Observer Design by Ackermann’s Formula

Transfer Function,
Y (s) 1
G (s) = = 2
U(s) s
Desired Characteristic Equation,

p(λ) = λ2 + 16λ + 100

Defining the States, x1 = y , x2 = ẏ

State-Space Model,    
0 1 0
ẋ = x+ u
0 0 1
Controllability Matrix,
 
  1 0
Qo = B AB =
2 3
 
1 0
Qo−1 =
−2/3 1/3
Control Systems - II Session: 2024-2025
Observer Design by Ackermann’s Formula

 2      
2 3 2 3 1 0 133 66
p(A) = + 16 + 100 =
−1 4 −1 4 0 1 −22 177

Observer Gain,
     
133 66 1 0 0 22
L= =
−22 177 −2/3 1/3 1 59
Control Systems - II Session: 2024-2025
Integrated State Feedback Control and Observer Design

u(t) = −K x̂(t) (57)

Observer Equation,
˙
x̂(t) = Ax̂(t) + Bu(t) + L(y (t) = C x̂(t)) (58)

Using the Feedback Control Law,

˙
x̂(t) = (A − BK − LC )x̂(t) + Ly (t) (59)

Error Dynamics,
˙
ė(t) = ẋ(t)−x̂(t) = Ax(x)+Bu(t)−Ax̂(t)−Bu(t)−Ly (t)+LC x̂(t) (60)

Simplifying,
ė(t) = (A − LC )e(t) (61)
Control Systems - II Session: 2024-2025
Integrated State Feedback Control and Observer Design

ẋ(t) = Ax(t) + Bu(t)

(62)
y (t) = Cx(t)

Feedback Control Law,

u(t) = −K x̂ (63)
Closed-loop System Dynamics,

ẋ(t) = Ax(t) + Bu(t) = Ax(t) − BK x̂(t)

(64)
ẋ(t) = (A − BK )x(t) + BKe(t)

In matrix-vector form,
    
ẋ(t) (A − BK ) BK x(t)
= (65)
ė(t) 0 (A − LC ) e(t)
Control Systems - II Session: 2024-2025
Integrated State Feedback Control and Observer Design

Characteristic Equation,

det(λI − (A − BK ))det(λI − (A − LC )) = 0 (66)

If the Eigenvalues of (λI − (A − BK )) as well as the Eigenvalues of

(λI − (A − LC )) lie in the left half s-plane, then, the overall System
is Stable.
Both the State Feedback Law and the Observer can be designed
independently → Separation Principle.
Control Systems - II Session: 2024-2025
Reduced-Order Observer

Significance of Reduced-Order Observer Design

Full-Order Observer provides those States also which are directly

available by measurement.
Observer Order and Complexity can be reduced by designing
Reduced-Order Observer.
The Reduced-Order Observer reduces the Order of the Observer by
the number of measured Outputs.
Control Systems - II Session: 2024-2025
Reduced-Order Observer

Assumption → The first State x1 appears at the Output and it can

be measured directly.

x → Vector containing the remaining n − 1 States to be Estimated.

      
ẋ1 A11 A12 x1 B
= + 1 u
ẋ A21 A22 x B2
  (67)
  x1
y= 1 0
x

The dynamics of the unmeasured States,

ẋ = A22 x + A21 x1 + B2 u (68)

Here, the rightmost two terms are competely known.

Control Systems - II Session: 2024-2025
Reduced-Order Observer

The dynamics of the measured State,

ẋ1 = ẏ = A11 y + A12 x + B1 u

(69)
ẏ − A11 y − B1 u = A12 x

Here, the left side terms are completely known.

The last Equation has the same relationship to the State, x as the
 >
original Equation to the entire States, x1 x .
Therefore, the following substitutions can be established:
 >
x1 x ←x
A ← A22
Bu ← A21 y + B2 u
y ← ẏ − A11 y − B1 u
C ← A12
Control Systems - II Session: 2024-2025
Reduced-Order Observer

Reduced-Order Observer,

x̂˙ = A22 x̂ + A21 y + B2 u + L(ẏ − A11 y − B1 u − A12 x̂) (70)

Defining the Estimation Error,

x̄ = x − x̂ (71)

Error Dynamics,
x̄˙ = (A22 − LA12 )x̄ (72)
Again, the gain L can be designed such that (A22 − LA12 ) is Hurwitz
matrix.
Control Systems - II Session: 2024-2025
Reduced-Order Observer

The Reduced-Order Observer can be expressed as,

x̂˙ = (A22 − LA12 )x̂ + (A21 − LA11 )y + (B2 − LB1 )u + Lẏ (73)

In the above Equation, the information of the derivative of y

is required.
Differentiation enhances noises which may be present in
Output measurement, y .

Defining new State,

ẑ = x̂ − Ly
The Reduced-Order Observer can be implemented as,

ẑ˙ = (A22 −LA12 )(ẑ +Ly )+(A21 −LA11 )y +(B2 −LB1 )u +Lẏ (74)
Control Systems - II Session: 2024-2025
Reduced-Order Observer

The Simple Pendulum System can be approximated by,

      
ẋ1 0 1 x1 0
= + u
ẋ2 −ω02 0 x2 1
  (75)
  x1
y= 1 0
x2
(A22 − LA12 ) = 0 − L = −L
Characteristic Equation,
s − (−L) = 0 =⇒ s + L = 0
Desired Characteristic Equation,
s + 10ω0 = 0
Comparing,
L = 10ω0
Reduced-Order Observer,
ẑ˙ = −10ω0 (ẑ − 10ω0 y ) − ω02 y + u
Control Systems - II Session: 2024-2025
PID Control

Defining Error,
e(t) = yr (t) − y (t)

PID Control Law,

Z t
kp e(t)
u(t) = kp e(t) + e(τ )dτ + kp τd (76)
τi 0 dt

where,
y (t) → Actual Output
yr (t) → Desired Output
kc → Proportional Gain
τi → Integral Time
τd → Derivative Time

Transfer Function,
U(s) kc
Gc (s) = = kc + + kc τd s
Yr (s) − Y (s) τi s
Control Systems - II Session: 2024-2025
Linearization

Consider a general Non-Linear System,

ẋ = f (x) (77)

Equilibrium Point
The Equilibrium Point is the point in the State Space at which the
derivatives of all the States are zero.
Once at the Equilibrium Point, the System trajectories continue to
remain at the point if left undisturbed.

At the Equilibrium Point,

x˙e = 0 = f (xe ) (78)

Control Systems - II Session: 2024-2025
Linearization

Linearization

Systems with Non-Linear components can be approximated as

Linear Systems.
Linearization is performed for small-signal inputs about the
steady-state solution when the small-signal input is 0.
This steady-state solution is the Equilibrium Point.
Control Systems - II Session: 2024-2025
Linearization

Linearized System,
ẋ = Ax (79)
At the Equilibrium Point,
Axe = 0 (80)

xe = 0 is the only solution provided |A| =

6 0.
If all the Eigenvalues of the System are different from 0, then, the
origin is the only Equilibrium Point.
Control Systems - II Session: 2024-2025
Linearization

For a non-zero Equilibrium Point, it is convenient to shift the origin of

co-ordinates to xe by defining new State variables,

x̄ = x − xe (81)

The same System may be represented by

x̄˙ = f (x̄) (82)

with the Equilibrium Point at x̄ = 0

Control Systems - II Session: 2024-2025
Linearization

Figure: Linearization

Non-linear System operating at point A[x0 , f (x0 )].

Small changes in input, δx = x − x0 .
Small changes in output, δf (x) = f (x) − f (x0 ).
δf (x) ≈ mδx (83)
f (x) ≈ f (x0 ) + m(x − x0 ) ≈ f (x0 ) + mδx (84)
Control Systems - II Session: 2024-2025
Linearization

Taylor’s Series Expansion

df (x − x0 ) d 2 f (x − x0 )2
f (x) = f (x0 ) + + 2 + · · · (85)
dx x=x0 1! dx x=x0 2!

Neglecting the higher-order terms,

df
f (x) ≈ f (x0 ) + (x − x0 ) (86)
dx x=x0
Control Systems - II Session: 2024-2025
Linearization

If the function depends on several variables, then,

y (t) = g (x1 (t), x2 (t), ..., xn (t)) (87)

∂g
y (t) = g (x10 , x20 , ..., x1n0 ) + (x1 (t) − x10 )
∂x1 x(t)=x0
∂g
+ (x2 (t) − x20 ) + · · · (88)
∂x2 x(t)=x0
∂g
···+ (xn (t) − xn0 )
∂xn x(t)=x0

where,
x0 → Operating Point
Control Systems - II Session: 2024-2025
Linearization

Example

f (x) = 5 cos x
df
= −5 sin x
dx
df
= −5
dx x= π
2

f (π/2) = 5 cos π/2 = 0 Figure: Linearization

fl (x) = −5x
Control Systems - II Session: 2024-2025
Linearization

Example

d 2x dx
2
+ 1.1 + cos x = 0
dt dt
d 2 (x + π4 ) d(x + π4 )
 
π
+ 1.1 + cos x + =0
dt 2 dt 4
   
π π d cos x
cos x + = cos − x
4 4 dx x= π
4
     
π π π
=⇒ cos x + = cos − sin x
4 4 4
d 2x dx 1 1
2
+ 1.1 − √ x = −√
dt dt 2 2
Control Systems - II Session: 2024-2025
Linearization

Consider the Non-Linear System in State-Space form,

 
f1 (x, u)
  f2 (x, u)
 
ẋ · 
ẋ = 1 = 

 = f (x, u) (89)
ẋ2  · 
 
 · 
fn (x, u)
The Linearized State-Space Model is,
ẋ = Ax + Bu (90)
where,
 ∂f1 ∂f1 ∂f1   ∂f1 
∂x1 ∂x2 · · · ∂xn ∂u
 ∂f2 ∂f2 ∂f2   ∂f2 
 ∂x ∂x2 · · · ∂xn 
  ∂u 
 1  
A= · · · · · ·  , B= · 
   
   
 · · · · · ·   · 
   
∂fn ∂fn ∂fn ∂fn
∂x1 ∂x2 · · · ∂xn (x0 ,u0 ) ∂u (x0 ,u0 )
Control Systems - II Session: 2024-2025
Linearization

The Equation of Motion of Simple Pendulum System is,

g Tc
θ̈ + sin θ =
l ml 2
where,
θ → Angular Deflection
m → Mass of the String
l → Length of the String
Tc → Input Torque

Defining the States,

x1 = θ, x2 = θ̇

The State-Space model is,

     
ẋ1 x2 f1 (x, u)
= = = f (x, u)
ẋ2 −ω02 sin x1 + u f2 (x, u)

where, p
ω0 = g /l, u = Tc /ml 2
Control Systems - II Session: 2024-2025
Linearization

Considering nominal Input Torque, u0 = 0

The State-Space model becomes,

ẋ1 = x2
g
ẋ2 = − sin θ
l
At the Equilibrium Point,

ẋ1 = 0 = x2
g
ẋ2 = 0 = − sin θ0
l
sin θ0 = 0 =⇒ θ0 = 0, π
Equilibrium Point,  
θ0
x0 = , u0 = 0
0
Control Systems - II Session: 2024-2025
Linearization

" ∂f ∂f1
#
1  
∂x1 ∂x2 0 1
A= =
∂f2 ∂f2 −ω02 cos θ0 0
∂x1 ∂x2 (x0 ,u0 )
" ∂f #
1
 
∂u 0
B= =
∂f2 1
∂u (x0 ,u0 )

Linearized Model in State-Space form,

      
ẋ1 0 1 x1 0
= + u
ẋ2 −ω02 cos θ0 0 x2 1
 
  x1
y= 1 0
x2