Inverted Pendulum System

Fall 2014

Output Feedback vs. State

Space Approach for Tracking

2014 Fall Inverted Pendulum Tracking

Inverted Pendulum System

x
θ

L/2

L
cos θ
L / 2 mg 2

F M

Fig 1. Inverted Pendulum System

Free Body Diagram : Rod & Cart

θ
About bar having mass ' m '

∑ F= x R=
x m 
xm  (1)

(a)
mg
L
2
cos θ ∑F y = Ry − mg = mym  (2)
Rx L L
Ry ∑ M =R x
2
cos θ + Ry sin θ = Iθ
2
 (3)

Ry
Rx

F M About cart having mass ' M '

(b)

∑F x =F − Rx =M 
x  (4)

Fig 2.
(a) free-body diagram of rod
(b) free-body diagram of cart
Variable Relation

L
xm= x − sin θ
2
y L
θ ym = cos θ
2
L
sin θ
2

xm L
xm= x − θ cos θ
L
cos θ
2
2 L
y m = − θ sin θ
x 2

x
L L
Fig 3. Relation between two center of gravities
 x + θ 2 sin θ − θ cos θ

xm =
2 2
L L
 − θ 2 cos θ − θsin θ
ym =
2 2
Equations & Desired State

From above relations, equations (1) ~ (4) can be rewritten as follow

 L L 
(1) Rx = x + θ 2 sin θ − θ cos θ 
m  
 2 2 
Initial condition
 L L 
(2) Ry = mg + m  − θ 2 cos θ − θsin θ  θ ( 0 ) = θ0
 2 2 
2 x ( 0) = 0
L L L
(3) Iθ = m 
x cos θ − m   θ + mg sin θ  (5)
2 2 2

(4) ( M + m ) x + m
2
(
L 2
)
θ sin θ − θ cos θ =
F  (6)
Equations & Desired State

From equation (5)

 
1  L   
2
L 

=
x   I + m    θ − mg sin θ 
L
m cos θ    2   2 
2
put this into (6) and arrange

1  2 L 
θ −m ( L / 2 ) sin (θ ) cos (θ ) θ 2 + ( M + m)mg sin (θ ) + m ( L / 2 ) cos (θ ) F 
2

(
( M + m) I + m ( L / 2 )
2
) 
− m 2 ( L / 2 ) cos 2 (θ ) 
2
2 

and from this, 

x also can be


1  m 2 ( L / 2 )2 sin (θ ) cos (θ ) g − m ( I + m( L / 2) 2 ) ( L / 2) sin(θ )θ 2 + ( I + m( L / 2) 2 ) F 
( )
x
( M + m) I + m ( L / 2 ) − m 2 ( L / 2 ) cos 2 (θ )  
2 2
1. Nonlinear System Control

Nonlinear System + Angle(θ ) and Angle Rate(θ ) Controller

x = θ θ x
T
x 

0 
ydes =  
0  I. 2nd Oscillation Controller θ 
= =  
II. P Controller Nonlinear y Cx
 x
III. PD Controller u System x

θ θ Not Fullstate Feedback

Fig 4. Block diagram for nonlinear system control

θ Control : by Force control

Desired state is that θ goes to zero at t → ∞

lim θ ( t ) = 0
(We do not care about x ( t ) ) t →∞

1  2  2 + ( M + m)mg L sin (θ ) + m ( L / 2 ) cos (θ ) F 

θ − ( ) (θ ) (θ ) θ
2

( )  m L / 2 sin cos 
( M + m) I + m ( L / 2 ) − m 2 ( L / 2 ) cos 2 (θ ) 
2 2
2

From above equation, we want to make θ → 0 by controlling appropriate F (t )

F is a control input u ( t ) , F (t ) = u ( t )

Three ways of control using force are introduced

I Make above equation about θlike θ =

−2ζωnθ − ωn2θ

II u ( t ) = F = − Kθ ( t )

III u (t ) =− Kθ ( t ) − Cθ
F=
θ Control : by Force control

1  2  2 + ( M + m)mg L sin (θ ) + m ( L / 2 ) cos (θ ) F 

θ − ( ) (θ ) (θ ) θ
2

( )  m L / 2 sin cos 
( M + m) I + m ( L / 2 ) − m 2 ( L / 2 ) cos 2 (θ ) 
2 2
2

( M + m)  L 
2 ( M + m) L
=
L
 I + m    − m cos θ and B = sin θ
Let A
L  2  
L
cos θ
2
m cos θ 
2 m
2 2

Then above equation can be expressed as follows

1 L B 1
θ =− m θ 2 sin θ + ⋅ mg + F
A 2 A A
If θ satisfies the following equation,

θ =
−2ζωnθ − ωn2θ

L
∴ u ( t ) =F =− 2 A ⋅ ζωn ⋅ θ − A ⋅ ωn2 ⋅ θ − B ⋅ mg + m sin θ ⋅ θ 2
2
With appropriate values of ζ and ωn , θ can be zero at t → ∞
θ Control : by Force control

1 L B 1
From the equation θ =− m θ 2 sin θ + ⋅ mg + F
A 2 A A
We want to control by

II u (t ) =− Kθ ( t ) ( K > 0 )
F=

1 L B K
θ =− m θ 2 sin θ + ⋅ mg − θ
A 2 A A

In this case, θ goes to zero when t → ∞ ?

III u ( t ) =− Kθ ( t ) − Cθ ( K > 0, C > 0 )

F=

θ =
1 L
− m θ 2 sin θ +
A 2
B
A
⋅ mg − (
1
A
Kθ + Cθ )
In this case, θ goes to zero when t → ∞ ?
Simulation : Force Control

Simulation
θ ( 0 ) = 0.1rad x ( 0) = 0 m ζ = 0.5 K = 30
Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2 ωn = 0.7 rad / s C = 15

I Simulation Results Plots ㅡ : ζ , ωn control --- : P control --- : PD control

0 1

-20 0

dx/dt [m]
x [m]

-40 -1

-60 -2

-80
0 10 20 30 40 50 -3
0 10 20 30 40 50
time [sec]
time [sec]
0.1 0.4

0.05 dθ/dt [rad] 0.2

θ [rad]

0 0

-0.05
-0.2
-0.1
0 10 20 30 40 50 -0.4
0 10 20 30 40 50
time [sec]
time [sec]
Fig 5. Simulation results of nonlinear system for three kinds of control
Simulation : Animation

Simulation
θ ( 0 ) = 0.1rad x ( 0) = 0 m ζ = 0.5 K = 30
Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2 ωn = 0.7 rad / s C = 15

II Force Plot and Simulation Results Animation

4
2nd Oscilation
2 P control
Force [N]

PD control
0 ㅡ : ζ , ωn control
-2
--- : P control
--- : PD control
-4
0 10 20 30 40 50
time [sec]

(a) 2nd Oscillation Control (b) P Control (c) PD Control

Fig 6. Simulation results Animation
After run simulation model

Open 'RUN _ file.m ' and execute

⇒ Four graphs (θ vs t , θ vs t , x vs t , x vs t )
&
Animation

Fig 7. Screenshot after run simulation

Nonlinear System

Y Equation of motions
x 2
L L L
θ Iθ = m 
x cos θ − m   θ + mg sin θ
2 2 2

( )
L/2 L
( M + m ) x + m θ2 sin θ − θ cos θ =F
L
cos θ
2
L / 2 mg 2

F M Control input : u ( t ) = F
X
θ ( 0 ) θ 0=
Initial Condition := , θ ( 0 ) 0,=
x ( 0 ) 0,=
x ( 0 ) 0

From combining above two equations

 x2  0
B 1 L 1    
θ = mg − m θ 2 sin θ + F  x1 
 B 1 L 2
mg − m θ sin θ  1
A A 2 A  x 
 A A 2   A
=x = 2
  +  0  F (t )
 x3  x4
 1 L B 1       

x= C  − m sin θθ 2 + mg + F  − g tan θ x
 4  C − m sin θθ 2 + mg − g tan θ   C 
  1 L B 
 A 2 A A    A 2    A 
 A 

( M + m)
 L 
2
L
, B=
( M + m)
L
sin θ
I + m ( L / 2)
2

Where =
A  I + m    − m cos θ C=
L  2   L m ( L / 2 ) cos θ
m cos θ  2 m cos θ 2
2 2
Angle and Angle Rate Controller

Control Strategy
− How to control u (t ) =
F (t ) ?
− Desired state when t → ∞ is θ → 0

i ) Applying a assumption that θ satisfies 2nd oscillation, bellow equation comes

B 1 L 2 1
θ =mg − −2ζωnθ − ωn2θ
m θ sin θ + F =
A A 2 A

In this case, control input becomes

u (t ) =F (t ) = ( ) L
A −2ζωnθ − ωn2θ − Bmg + m sin θ ⋅ θ 2
2

ii ) P control
u ( t ) = F ( t ) = − Kθ

iii ) PD control
u (t ) =
F (t ) =
− Kθ − Cθ
2. Linear System Control

Full State Feedback Controller

Linear System +
I. 0  Reference Tracking Controller
ydes =  
0 
II.  0 
ydes =  
 xdes  I. Full State Feedback Controller
u = -Kx θ 
= =  
Linear y Cx
 x
II. Reference Tracking Controller u System x
u = -Kx+r

x = θ θ x
T
x 

Fig 8. Block diagram for linear system control

Linear Model

By small angle assumption, sin (θ ) ≈ θ , cos (θ ) ≈ 1, θ 2 ≈ 0

 x2  0
 x1   B 1 L   
 x   mg − m θ 2 sin θ  1
 A A 2   A
then, =x = 2
  +  0  F (t ) becomes as follows
 x3  x4
     
 x4    1 L  B   C 
C  − A m 2 sin θθ + A mg  − g tan θ   A 
2

 

 0 1 0 0  0 
   
 mg ( L / 2 )  m ( L / 2) / ( M + m)

 x1  0 0 0  
 x   I + m ( L / 2 ) − m ( L / 2 ) / ( M + m)
2 2 2
  I + m ( L / 2 ) − m ( L / 2 ) / ( M + m) 
2 2 2
 
=x = 2 
  x+  F (t )
 x3  0 0 0 1  0 
 
  

 4 
x 

I + (m ( L / 2 )
2
)  
− 1 g 0 0 0 

 I + m ( L / 2 )
2
(
/ ( M )
+ m ) 

 
 I + m ( L / 2 ) − m ( L / 2 ) / ( M + m) 

( ) ( )
2 2
2
 
 I + m L / 2
2
− m 2
L / 2
2
/ ( M + m ) 
  

x1 x2 x3 x4 ]T θ θ x x 
T
=
where x [ =

( M + m )  I + m  L 2  − m L B=
( M + m) L θ I + m ( L / 2)
2
Where A=     , C=
m
L
  2   2 m
L 2 m ( L / 2)
2 2
Linear Model : Feedback Control

When a system is defined as x= Ax + Bu , we want to control x to make x(∞)= 0

Let u = − Kx
Then, x = Ax − Bu = ( A − BK ) x ⇒ x = e( A− BK )t

( x : n ×1 matrix, A : n × n matrix , B : n ×1 matrix, K :1× n matrix )

Appropriate gain K makes the system stable

θ θ x x  , K = [ k1 k2 k3 k4 ]
=where x [ =x1 x2 x3 x4 ]T
   
T

If eigenvalues ( poles ) of matrix A − BK have negative − real part ,

then x goes to zero at t → ∞

The control methods , which were introduced at the 'Nonlinear system' part,

! were failed to control state 'x' to be zero at t → ∞

On the otherhand, linear feedback control make the system to be stable about all state

Command 'acker ' can be used to get desired gain K

Simulation : Linear Feedback Control

θ ( 0 ) = 0.1rad x ( 0) = 0 m
Simulation
=K [70.5, 22.7, − 0.3, − 0.6483]
Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2

Simulation Results Plots

0.15 0.05

0.1 0

dθ/dt [rad/s]
θ [rad]

-0.05
0.05
-0.1
0
-0.15
-0.05 -0.2
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]
1 0.4
0.5 0.2

0
dx/dt [m/s] 0
x [m]

-0.5 -0.2
-1 -0.4
-1.5 -0.6
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]
Fig 9. Simulation results of linear system for tracking control
Simulation : Linear Feedback Control - Animation

θ ( 0 ) = 0.1rad x ( 0) = 0 m
Simulation
=K [70.5, 22.7, − 0.3, − 0.6483]
Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2

Force Plot and Simulation Results Animation

6
Force [N]

-2
0 20 40 60 80 100
time [sec]

Fig 10. Feedback Control Input and Simulation results Animation

Linear Model : Tracking Problem

We discussed linear feedback control.

This make it possible that all the states go to zero at t → ∞.
But what if we want to make the states converge to specific values ?

Q ) How to make the states converge to nonzero values

Reference Input Tracking

Introduce reference input u=

− Kx + r

Let xss and uss as state x and input u repectively at steady state
Then, u =uss − K ( x − xss )

x Ax + Bu
=
x1 x2 x3 x4 ]T θ θ x x  ,
T
if the system is like =where x [ =
=y Cx + Du
=0 Ax ss + Buss
At the steady state, this system becomes  (*) ( At steady state, x = 0 )
=
yss Cx ss + Duss
Linear Model : Tracking Problem

We want to make yss = rss for any value of rss

x ss = N x rss
To do this, asssume that and put these equations to (*)
uss = N u rss

= 0 AN x rss + BN u rss
Then
= rss CN x rss + DN u rss
 A B   Nx  0
It can be also written as a matrix form like C D   N  =  1 
  u  

A B
Assume that the inverse of   is exists, then this equation can be solved for N x and N u
C D 
−1
 Nx   A B  0
 N  = C D   1 
 u    
From above relation
Nu r − K ( x − N x r ) =
u= − Kx + ( N u + K N x ) r
where =
N Nu + K N x
=
− Kx + Nr
Simulation : Reference Tracking Control

Simulation
θ ( 0 ) = 0.1rad x ( 0) = 0 m =K [70.5, 22.7, − 0.3, − 0.6483]
Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2 xss= rss= 5

Simulation Results Plots

0.15 0.1

0.1 0

dθ/dt [rad/s]
θ [rad]

0.05 -0.1

0 -0.2

-0.05 -0.3
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]

8 1.5

6 1
dx/dt [m/s]
4 0.5
x [m]

2 0

0 -0.5

-2 -1
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]

Fig 11. Simulation results of linear system for tracking control

Simulation : Reference Tracking Control - Animation

Simulation
θ ( 0 ) = 0.1rad x ( 0) = 0 m =K [70.5, 22.7, − 0.3, − 0.6483]
Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2 xss= rss= 5

Force Plot and Simulation Results Animation

5
Force [N]

-5

-10
0 20 40 60 80 100
time [sec]
Fig 12. Reference Tracking Input and Simulation results Animation
3. Nonlinear System with Linear Controller

Full State Feedback Controller

Nonlinear System +
I. 0  Reference Tracking Controller
ydes =  
0 
II.  0 
ydes =  
 xdes  I. Full State Feedback Controller
u = -Kx θ 
= =  
Nonlinear y Cx
 x
II. Reference Tracking Controller u System x
u = -Kx+r

x = θ θ x
T
x 

Fig 8. Block diagram for nonlinear system with linear controller

Nonlinear System

Y Equation of motions
x 2
L L L
θ Iθ = m 
x cos θ − m   θ + mg sin θ
2 2 2

( )
L/2 L
( M + m ) x + m θ2 sin θ − θ cos θ =F
L
cos θ
2
L / 2 mg 2

F M Control input : u ( t ) = F
X
θ ( 0 ) θ 0=
Initial Condition := , θ ( 0 ) 0,=
x ( 0 ) 0,=
x ( 0 ) 0

From combining above two equations

 x2  0
B 1 L 1    
θ = mg − m θ 2 sin θ + F  x1 
 B 1 L 2
mg − m θ sin θ  1
A A 2 A  x 
 A A 2   A
=x = 2
  +  0  F (t )
 x3  x4
 1 L B 1       

x= C  − m sin θθ 2 + mg + F  − g tan θ x
 4  C − m sin θθ 2 + mg − g tan θ   C 
  1 L B 
 A 2 A A    A 2    A 
 A 

( M + m)
 L 
2
L
, B=
( M + m)
L
sin θ
I + m ( L / 2)
2

Where =
A  I + m    − m cos θ C=
L  2   L m ( L / 2 ) cos θ
m cos θ  2 m cos θ 2
2 2
Linear Feedback Control

When a system is defined as x= Ax + Bu , we want to control x to make x(∞)= 0

Let u = − Kx
Then, x = Ax − Bu = ( A − BK ) x ⇒ x = e( A− BK )t

( x : n ×1 matrix, A : n × n matrix , B : n ×1 matrix, K :1× n matrix )

Appropriate gain K makes the system stable

θ θ x x  , K = [ k1 k2 k3 k4 ]
=where x [ =x1 x2 x3 x4 ]T
   
T

If eigenvalues ( poles ) of matrix A − BK have negative − real part ,

then x goes to zero at t → ∞

The control methods , which were introduced at the 'Nonlinear system' part,

! were failed to control state 'x' to be zero at t → ∞

On the otherhand, linear feedback control make the system to be stable about all state

Command 'acker ' can be used to get desired gain K

Nonlinear System vs Linear System with Linear Feedback Control

θ ( 0 ) = 0.1rad x ( 0) = 0 m
Simulation
=K [70.5, 22.7, − 0.3, − 0.6483]
Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2

Simulation Results Plots

0.15 0.05
Linear System
0
0.1

dθ/dt [rad/s]
Nonlinear System
θ [rad]

-0.05
0.05
-0.1
0
-0.15

-0.05 -0.2
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]
1 0.05

0.5 0

0 dθ/dt [rad/s] -0.05

x [m]

-0.5 -0.1

-1 -0.15

-1.5 -0.2
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]
Fig 9. Simulation results for linear Feedback control
Nonlinear System vs Linear System with Linear Feedback Control

θ ( 0 ) = 0.1rad x ( 0) = 0 m
Simulation
=K [70.5, 22.7, − 0.3, − 0.6483]
Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2

Force Plot and Simulation Results Animation

8
Linear System
6
Nonlinear System
Force [N]

-2
0 20 40 60 80 100
time [sec]
8
Fig 10. Feedback Control Input (Nonlinear system)
Linear System and Simulation results Animation
6
Nonlinear System
Force [N]

-2
0 0.5 1 1.5 2 2.5 3
time [sec]
Nonlinear System vs Linear System with Linear Feedback Control

θ ( 0 ) = 1.57 rad x ( 0) = 0 m
Simulation
=K [70.5, 22.7, − 0.3, − 0.6483]
Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2

Simulation Results Plots

10 40
Linear System
5 20

dθ/dt [rad/s]
Nonlinear System
θ [rad]

0 0

-5 -20

-10 -40
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]
100 100

50
0 dx/dt [m/s] 0
x [m]

-50
-100
-100

-200 -150
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]
Fig 11. Simulation results for linear feedback control
Nonlinear System vs Linear System with Linear Feedback Control

θ ( 0 ) = 1.57 rad x ( 0) = 0 m
Simulation
=K [70.5, 22.7, − 0.3, − 0.6483]
Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2

Force Plot and Simulation Results Animation

2000
Linear System
1000 Nonlinear System
Force [N]

-1000

-2000
0 20 40 60 80 100
time [sec]
2000
Linear System Fig 12. Feedback Control Input (Nonlinear system)
1000 Nonlinear System and Simulation results Animation
Force [N]

-1000

-2000
0 0.5 1 1.5 2 2.5 3
time [sec]
Nonlinear System with Linear Feedback Control

θ ( 0 ) = 2.44 rad x ( 0) = 0 m
Simulation
=K [70.5, 22.7, − 0.3, − 0.6483]
Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2

Simulation Results Plots

15 40

10 20

dθ/dt [rad/s]
θ [rad]

5 0

0 -20

-5 Linear System -40

Nonlinear System
-10 -60
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]
2000 300

1500 200

1000 dx/dt [m/s] 100

x [m]

500 0

0 -100

-500 -200
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]
Fig 13. Simulation results for linear feedback control
Nonlinear System with Linear Feedback Control - Animation

θ ( 0 ) = 2.44 rad x ( 0) = 0 m
Simulation
=K [70.5, 22.7, − 0.3, − 0.6483]
Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2

Force Plot and Simulation Results Animation

2000
Linear System
1000 Nonlinear System
Force [N]

-1000

-2000
0 20 40 60 80 100
time [sec]

Fig 14. Feedback Control Input (Nonlinear system) and Simulation results Animation
Tracking Problem

We discussed linear feedback control.

This make it possible that all the states go to zero at t → ∞.
But what if we want to make the states converge to specific values ?

Q ) How to make the states converge to nonzero values

Reference Input Tracking

Introduce reference input u=

− Kx + r

Let xss and uss as state x and input u repectively at steady state
Then, u =uss − K ( x − xss )

x Ax + Bu
=
x1 x2 x3 x4 ]T θ θ x x  ,
T
if the system is like =where x [ =
=y Cx + Du
=0 Ax ss + Buss
At the steady state, this system becomes  (*) ( At steady state, x = 0 )
=
yss Cx ss + Duss
Tracking Problem

We want to make yss = rss for any value of rss

x ss = N x rss
To do this, asssume that and put these equations to (*)
uss = N u rss

= 0 AN x rss + BN u rss
Then
= rss CN x rss + DN u rss
 A B   Nx  0
It can be also written as a matrix form like C D   N  =  1 
  u  

A B
Assume that the inverse of   is exists, then this equation can be solved for N x and N u
C D 
−1
 Nx   A B  0
 N  = C D   1 
 u    
From above relation
Nu r − K ( x − N x r ) =
u= − Kx + ( N u + K N x ) r
where =
N Nu + K N x
=
− Kx + Nr
Nonlinear System vs Linear System with Reference Tracking Control

Simulation
θ ( 0 ) = 0.1rad x ( 0) = 0 m =K [70.5, 22.7, − 0.3, − 0.6483]
Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2 xss= rss= 5

Simulation Results Plots

0.15 0.1
Linear System
0.1 0

dθ/dt [rad/s]
Nonlinear System
θ [rad]

0.05 -0.1

0 -0.2

-0.05 -0.3
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]
8 1.5

6 1

4 dx/dt [m/s] 0.5

x [m]

2 0

0 -0.5

-2 -1
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]
Fig 15. Simulation results for tracking control
Nonlinear System vs Linear System with Reference Tracking Control

Simulation
θ ( 0 ) = 0.1rad x ( 0) = 0 m =K [70.5, 22.7, − 0.3, − 0.6483]
Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2 xss= rss= 5

Force Plot and Simulation Results Animation

5
Linear System
Nonlinear System
Force [N]

-5

-10
0 20 40 60 80 100
time [sec]
5

Fig 16. Reference Tracking Input (Nonlinear system)

Force [N]

0 and Simulation results Animation

-5
Linear System
Nonlinear System
-10
0 0.5 1 1.5 2 2.5 3
time [sec]
Nonlinear System vs Linear System with Reference Tracking Control

Simulation θ ( 0 ) = 1.57 rad x ( 0) = 0 m =K [70.5, 22.7, − 0.3, − 0.6483]

Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2 xss= rss= 5

Simulation Results Plots

10 40
Linear System
5 20

dθ/dt [rad/s]
Nonlinear System
θ [rad]

0 0

-5 -20

-10 -40
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]
100 100

50
0 dx/dt [m/s] 0
x [m]

-50
-100
-100

-200 -150
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]
Fig 17. Simulation results for tracking control
Nonlinear System vs Linear System with Reference Tracking Control

Simulation θ ( 0 ) = 1.57 rad x ( 0) = 0 m =K [70.5, 22.7, − 0.3, − 0.6483]

Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2 xss= rss= 5

Force Plot and Simulation Results Animation

2000
Linear System
1000 Nonlinear System
Force [N]

-1000

-2000
0 20 40 60 80 100
time [sec]
2000
Linear System
Fig 18. Reference Tracking Input (Nonlinear system)
Nonlinear System
1000 and Simulation results Animation
Force [N]

-1000

-2000
0 0.5 1 1.5 2 2.5 3
time [sec]
Nonlinear System vs Linear System with Reference Tracking Control

Simulation θ ( 0 ) = 3.05 rad x ( 0) = 0 m =K [70.5, 22.7, − 0.3, − 0.6483]

Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2 xss= rss= 5

Simulation Results Plots

10 40

5 20

dθ/dt [rad/s]
θ [rad]

0 0

-5 Linear System -20

Nonlinear System
-10 -40
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]
200 200

100
0 dx/dt [m/s]
x [m]

0
-200
-100

-400 -200
0 20 40 60 80 100 0 20 40 60 80 100
time [sec] time [sec]
Fig 19. Simulation results for tracking control
Nonlinear System vs Linear System with Reference Tracking Control

Simulation θ ( 0 ) = 3.05 rad x ( 0) = 0 m =K [70.5, 22.7, − 0.3, − 0.6483]

Condition θ ( 0 ) = 0 rad / s x ( 0 ) = 0 m / s 2 xss= rss= 5

Force Plot and Simulation Results Animation

2000
Linear System
1000 Nonlinear System
Force [N]

-1000

-2000
0 20 40 60 80 100
time [sec]

Fig 20. Reference Tracking Input (Nonlinear system)

and Simulation results Animation
Appendix. A
How to use

Open ‘RUN_file.m’

1. Simulation(시간, 초기조건 등) 환경 설정
2. Parameter(질량, 길이 등) 값 설정
3. Gain 값 설정
4. 실행
5. ‘Simulaton_file.m’이 실행된다.

‘Simulation_file.m’

5. Nonlinear Equation을 실시간으로 계산(By function ‘ODE4 ’)

6. State의 실시간 값들이 얻어짐(x,theta…)
7. State의 시간에 대한 그래프 출력
How to use – RUN_file.m

Open ‘RUN_file.m’ Simulation Setting

Global 변수 선언

시뮬레이션에 필요한 함수인 ode45에 쓰일 변수를 선언.

부록B에 ode45에 대해서 설명해두었다.

Simulation time 설정
t0 : 초기 시간(=0)
tf : simulation 종료 시간
del_t : sampling time 설정

Initial Condition 설정
x0 : Cart의 초기 위치
d_x0 : Cart의 초기 속도
theta_0 : Rod의 초기 각도
d_theta_0 : Rod의 초기 각속도
Desired X 좌표 설정
rss : 목표 도달 X 좌표

Dynamic Parameter 설정
m : Rod의 질량
M : Cart의 질량
L : Rod의 길이
I : Rod의 moment of inertia (자동으로 계산)
g : 중력가속도 (고정값)

Control Gain 설정
K : P gain 값 설정 (양수)
C : D gain 값 설정 (양수, 0일 시에는 P Control)
Zeta : 댐핑 값 설정
wn : 공진주파수 설정
How to use – Simulation_file.m

▶ ‘RUN_file’ 을 실행시키면 자동으로 ‘Simulation_file.m ’이 실행된다.

ode45
Ode45라는 함수를 통해 nonlinear equation을 실시
간으로 풀어낼 수 있다.(부록B에 ode45에 대해 설명)

State의 실시간 값들이 얻어짐

앞서 설정한 simulation time동안의 매 순간의 data
값(카트 위치, 바의 각도 등)들이 ode45 함수의 결과
로 얻어진다.

결과의 시간에 대한 그래프 출력

이렇게 구해진 실시간 데이터 결과들을 시간에 대한
그래프로 출력한다.
Appendix. B
부록 : ode45

▶ ODE45 ?

MATALB에 내장돼있는 함수 & 미분방정식을 실시간으로 풀어주는 함수

이 때 exact solution을 구하는 것이 아닌, 수치해석 방법으로 실시간 해를 구한다.
Runge-Kutta method를 사용하는 numerical solver
내가 풀고자 하는 미분방정식을 풀어주는 함수를 만든다.

▶ Why ode45?

선형 뿐 아니라 비선형 미분방정식도 풀 수 있으며

사용방법 및 수식작성이 직관적이다.

▶ How to use?

함수 이름 작성, 변수 선언
(이 때 함수 이름은 저장 시에도 동일한 이름으로 저장해야 한다.)

x = f ( x ) 꼴 그대로 수식으로 옮긴다.

부록 : ode45

x
m
Example
mx + cx += ( 0 ) x0 , x=
kx 0, x= ( 0 ) x0 k c

1. 함수의 이름을 정하고 변수를 차례로 적어준다.(이때는 시간과 state x)

미분 term을 나타내는 변수, state를 한번 미분한 것을 의미한다. 즉, 𝒙𝒙̇ 를 의미

함수이름(저장 시 동일한 이름으로 저장해야한다.)

2. 미분방정식에 쓰이는 global 변수(state 제외)를 선언한다.

이 함수에서 이 변수들이 쓰이기 때문에 이 함수가 돌아가기 전,

이 변수들이 workspace 상에 선언되어 있어야 한다.

3. x = f ( x ) 꼴로 식을 만들어주고 code에 그대로 옮겨 적는다.

x = x dx가 (2x1)행렬임을
 x   x1 
1 2
먼저 선언해야 한다.
=
x =  x  , c k
   2  x2 =

x − x2 − x1
m m
부록 : ode45

▶ 문제의 미분방정식을 푸는 ode45를 MATLAB에 작성하면 다음과 같다.

Example
mx + cx += ( 0 ) x0 , x=
kx 0, x= ( 0 ) x0

▶ 그럼 이 때 다음과 같이, 앞에서 만든 ode45를 이용한 m-file을 작성하여 실행하면 다음과 같다.

10
Displacement
Velocity
5

X
-5

-10
0 2 4 6 8 10
t

사용할 함수 이름 t의 초기, 말기 값 State의 초기 값(x1, x2 차례로)