7

201
NonLinear Control Engineering
Doctorado y Maestría Ingeniería Mecánica.

rás
Feedback Linearization Part I

Bor
Carlos Borrás P. Ph.D., MSc., Sp.
los
Escuela de Ingeniería Mecánica
Universidad Industrial de Santander
Car

October 22, 2020

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 1 / 72
 Índice

7
201
1 Feedback Linearization: General Idea
Input State Feedback Linearization

rás
Example 1 of: Input State Feedback Linearization
Example 2 of: Input State Feedback Linearization

2 Feedback Linearization Theory Bor

los
Input- Output Linearization
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 2 / 72
 Feedback Linearization- Slotine's Book

7
The approach involves coming up with a transformation of the nonlinear

201
system into an equivalent linear system through a change of variables and a
suitable control input.
The idea is to algebraically transform a NLSD into a linear one

rás
(fully or partially)

Bor
One of the most expeditious alternatives for linearization is to use feedback
of the state variables. This kind of control strategy is divided into two
types. One for linearize the input states; that is, to obtain linear
los
responses in variables system status and the other for the output
variables; that is, a linear behavior of outputs with respect to inputs
Car

regardless of Probably non-linear behavior of the states.

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 3 / 72
 Feedback Linearization- Slotine's Book

7
201
Both alternatives they need to know the plant model and deliver design

rás
options that They allow dierent advantages in both cases.
Before analyzing in detail the dierent control alternatives via exact

Bor
feedback, We are going to reviewing examples of each of these to
familiarize yourself with the control structures.
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 4 / 72
 Fundamentals: Basic concepts.

7
A Feedback Linearization and the Canonical Form.

201
The fundamental idea of linearization via feedback is the cancellation
of the non-linearities of a system by redening the entrance to it. For
example, in the case of the Tank (pond) in which the area is a

rás
function of height, you have the following model. (Slotine chapter 6)

Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 5 / 72
 Example. Linearization Tank problem

7
The fundamental idea of linearization via feedback is the cancellation of the

201
non-linearities of a system by redening the entrance to it. For example, in
the case of the Tank in which the area is a function of height, you have the
following model:

rás
nR o √
1 dtd 0h A(z)dz = u − a 2gh

Bor
2 So, we can symplied to:
√
3 A(h)ḣ = u − a 2gh
4 Therefore, we choose the control u as follow:
los
√
5 u = a 2gh + A(h)v
Car

6 With v is a new input control

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 6 / 72
 Example. Linearization Tank problem

7
201
7 dh(t)
dt = ḣ = v
8 Which is linear and therefore, allows you to choose v to achieve
additional control objectives. For example:

rás
9 v = β (hd − h) = ḣ
10 With β a positive constant and hd the height reference, So you have a

Bor
resulting dynamic given by:
11 ḣ + β h = β hd
12 Which indicates that h → hd when t → ∞. According to the previous
los
equations, the input is nally given by,
√ √
Car

13 u = a 2gh + A(h)β (hd − h), → u = a 2gh + A(h)β e

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 7 / 72
 Example. Linearization Tank problem

14 Which indicates that h → hd when t → ∞. According to the previous

7
201
equations, the input is nally given by,
√ √
15 u = a 2gh + A(h)β (hd − h). → u = a 2gh + A(h)β e
16 with error = e, e = hd − h. A more better version might consider

rás
that hd is known and variable time, so that the control input v can be
chosen as v = h˙d + β (hd − h), so that the resulting linear dynamics

Bor
would look like:
17 ḣ = h˙d + β (hd − h)
18 ė + β e = 0, where: → ė = ḣ − h˙d and e = hd − h
los
Where you clearly see that the error e → 0, when t → ∞. The application
of this redening the input control in order to cancel non-linearities, can be
Car

applied if the system has a model of the controllable canonical form, whose
general form for SISO systems is,
GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 8 / 72
 Example. Linearization Tank problem
Where you clearly see that the error e → 0, when t → ∞. The application
of this redening the input control in order to cancel non-linearities, can be

7
applied if the system has a model of the controllable canonical form, whose

201
general form for SISO systems is:
x (n) = f (x) + b(x)u

rás
Where u is the control input, and x is the output states,
X = x, ẋ, ẍ, .....x n−1 T = [x1 , x2 , x3 , x4 , .....xn ]T is the vector of states and
 

Bor
f (x) and b(x) are non-linear scalar functions of states. The above
representation can also be written as;
   
x1 x2
los

 x2 
  x3 

. .
Car

   
d  1
x) + v)
   
dt  .=
  . 
 →u= b(x ) (−f (

 . 
  . 

xn−1   xn  GRAPHS/UIS.p

xn f (x ) + b(u )
Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 9 / 72
 Example. Linearization Tank problem

7
x) + v)

201
1
u= b(x ) (−f (

This last representation allows us to see that if the control input u is

re-dened.

rás
with v a new entry, you get the resulting model.

Bor
xn = v

which corresponds to a linear model and therefore the linear control theory
can be used to achieve specic control objectives. For example, if the entry
los
is dened as:
Car

v = −a0 x − a1 ẋ − a2 ẍ − ....., −an−1 x n−1

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 10 / 72
 Example. Linearization Tank problem
v = −a0 x − a1 ẋ − a2 ẍ − ....., −an−1 x n−1

7
with a0 , a1 , a2 , ..., an−1 , chosen so that the polynomial

201
S n + an−1 S n−1 + .....a0 x = 0, has its roots in LPS, it leads to an
exponentially stable dynamic.
Therefore it implies that: x(t) → 0, when t → ∞

rás
When we need to follow a reference (tracking input) xd , a desired
trajectory, we can re-dened the input control as follow:

Bor
v = xdn + a0 e + a1 ė + ....... + an−1 e (n−1)
with e(t) = xd (t) − x(t), leads to the exponential convergence of the error;
los
meaning that: e(t) → 0, when t → ∞
Unfortunately, the models are not generally found in this controllable
Car

canonical form and by techniques are studied so much that allow to

transform an arbitrary system to this form. Like the linear systems, not all
systems can be transformed and this depends on their characteristicsGRAPHS/UIS.p
private individuals
Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 11 / 72
 Input State Linearization.
B Input States Linearization .
Consider the case of a system of an input of the form x = f (x, u) in

7
which you want to nd a new input in order to obtain a resulting

201
linear system. This is done in two steps, rst there is a transformation
of states z = z(x) and then an input transformation u = u(x, v ), so
that the original system becomes an linear time invariant equivalent of

rás
the form z = Az + bv, to then design v in order to obtain a desired
response, and use standard linear techniques such pole placement to

Bor
design v, please see Figure. los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 12 / 72
 Feedback Linearization Technique:Input State
Linearization.

7
Key Idea: transform the Non Linear System into a fully or partially

201
linearized system, so that te feedback control techniques could applied.
Consider the system.
ẋ = f (x) + g (x)u

rás
y = h(x)
If there exist a state feedback control.

Bor
u = α(x) + β (x)V
and a change of state variable:
los
Z = T (x)
That it can transform the Non Linear system into a equivalente linear
Car

system then all classical control techniques can be applied to the resulting
closed loop system.
u(t) = −Kx(t)
GRAPHS/UIS.p

:
Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 13 / 72
 Feedback Linearization Technique: Input State
Linearization.

7
Key Idea: transform the Non Linear System into a fully or partially

201
linearized system, so that te feedback control techniques could applied.
That it can transform the Non Linear system into a equivalente linear
system then all classical control techniques can be applied to the

rás
resulting closed loop system.
u(t) = −Kx(t)

Bor
ẋ = (A − BK)x(t) ∴ x(t) = e (A−BK) x(0)
:
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 14 / 72
 Input State Linearization.
Consider a nonlinear form: ẋ = f(x, u)
Method:

7
First: Find a state transformation z = z(x), so that the NLSD is

201
transformed into an equivalent Linear Time Invriant dynamic, in the
form: ż = Az + bv
Second: Uses standar linear approach to desing v , (such as pole

rás
placement, LQR)
Example: Consider the system.

Bor
x˙1 = −2x1 + ax2 + sin(x1 )
x˙2 = −x2 ∗ cos(x1 ) + u ∗ cos(2x1 )
los
The system is very hard and it is not in canonical form. It has an
equilibrium point at (0, 0). A specic diculty is the nonlinear in the rts
Car

equation, which is can not be directly canceled by the control input u .

However, if we consider the new set of state variables.
z1 = x1 GRAPHS/UIS.p

z2 = ax2 + sen(x1 )
Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 15 / 72
 Input State Linearization.

7
201
z1 = x1
z2 = ax2 + sen(x1 )

rás
Then the new state equations are:
z˙1 = x˙1 : → z˙1 = −2z1 + z2

Bor
z˙2 = ax˙2 + cos(x1 ) ∗ x˙1 Then the new state equations are:
z˙1 = x˙1 : → z˙1 = −2z1 + z2
z˙2 = ax˙2 + cos(x1 ) ∗ x˙1 :→ z˙2 =
los
−2z1 ∗ cos(z1 ) + cos(z1 ) ∗ sen(z1 ) + u ∗ a ∗ cos(2z1 )
They share the same origin of x as well as z
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 16 / 72
 Input State Linearization.
If we consider the new set of state variables. Then the new state equations
are:

7
201
z˙1 = −2z1 + z2
z˙2 = −2z1 ∗ cos(z1 ) + cos(z1 ) ∗ sen(z1 ) + u ∗ a ∗ cos(2z1 )
The new state equations also have an equilibrium point at (0, 0). Now the

rás
nonlinearities can be canceled by control law of the form:
1

Bor
u= a∗cos(2z1 ) (v − cos(z1 ) ∗ sen(z1 ) + 2z1 ∗ cos(z1 ))
Also, we can show the control u as well as function of the original states:
1
(v − cos(x1 ) ∗ sen(x1 ) + 2x1 ∗ cos(x1 ))
los
u= a∗cos(2x1 )

Where v is an equivalente input to be designed, leading to a linear

Car

input-state relation:
z˙1 = −2z1 + z2
GRAPHS/UIS.p
z˙2 = v
Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 17 / 72
 Transformation: it is linear and controllable.

7
The new dynamic is linear and controllable, it is well known that the linear

201
state feedback control law (pole placement can be used)
v = −k1 ∗ z1 −k2 ∗z2
z

rás
v = − [k1 k2 ] 1
z2

Bor
For example: if we chose v = −2z2 , therefore:
Resulting in stable closed loop dynamics
z˙1 = −2z1 + z2
los
z˙2 = −2z2
k(λ I − A)k = 0
Car

Whose poles are both placed at -2.

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 18 / 72
 Input-State Linearization

7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 19 / 72
 Remarks about the control law

The control law u = a∗cos(1 2x1 ) (v − cos(x1 ) ∗ sen(x1 ) + 2x1 ∗ cos(x1 )),

7
201
its not well dened when x1 = π/4 ± kπ/2), → k = 1, 2, ...., When the
initial state is at such singularity points, the controller cannot bring
the system to the equilibrium point.

rás
This is a transformation of the state and the input trasnformation. Its
quite dierent from the Jacobian linearization.

Bor
The new state components (z1, z2) must be available. If not, it must
be computed using the original states.
We must know the parameters a and the model of the system, any
los
uncertainty in the model will cause error in the computation of both
new states z and the control u .
Car

Tracking control problem, can be considered. but the desired motion

must be exresed in terms of full new state. vector
GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 20 / 72
 Example 2
Considere the following system:

7
x˙1 = a · sin(x2 )

201
x˙2 = −x12 + u
It is dicult to nd u to cancel the Nonlinear term a · sin(x2 ), therefore, its
suggested to change variables as follow:

rás
z1 = x1

Bor
z2 = a · sin(x2 ) = x˙1
So,
los
z˙1 = z2
z˙2 = a · cos(x2 )(x˙2 ) = a · cos(x2 )(−x12 + u)
Car

Therefore we got:
2 1
u = x1 + v → u = α(x) + β (x)v , with 1(x)
β (x) = γ −GRAPHS/UIS.p
acos(x2 )
Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 21 / 72
 Example 2

7
z˙1 = z2

201
z˙2 = a · cos(x2 )(x˙2 ) = a · cos(x2 )(−x12 + u)
Therefore we got:

rás
2 1
u = x1 + v → u = α(x) + β (x)v , with β (x) = γ − 1(x)
acos(x2 )

Bor
−π/2 < x2 < π/2
The linear new variable system becomes as:
los
z˙1 = z2
z˙2 = v
Car

Now the system is linear: choose v as follow, to placed both poles.

v = −K1 · z1 − K2 · z2 → k(λ I − A)k = 0 GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 22 / 72
 Questions

7
201
rás
What Clases of nonlinear systems can be transformed into linear
systems.

Bor
How to nd the proper transformation for those which can?
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 23 / 72
 Feedback Linearization
A Continuously dierentiable Map T(x) is a dieormorphism if

7
T−1 (x) is continuously dierentiable. This is true if the Jacobian

201
Matrix ∂∂Tx is NonSingular ∀x ∈ D
T(x) is a global dieormorphism if and only of ∂∂Tx is NonSingular
∀x ∈ Rn , and T(x) is proper, meaning that: limkxk→∞ kT(x)k = ∞

rás
DEFINITION

Bor
A non Linear System:
ẋ = f(x) + g(x)u (1)
Where f : D → Rn ,
and g : D → Rnxp ,
are suciently smooth on a domain
los
D ⊂ R , is said to be Feedback Linearizable (or input-state
n

linearizable, if there exists a dieomorphism T : D → Rn such that

Car

Dz = T(D) contains the origin and the change of variables z = T (x)

transform the equation 1 into the form:
GRAPHS/UIS.p

ż = Az + Bγ(x)(u − α(x)) (2)

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 24 / 72
 Feedback Linearization

7
DEFINITION

201
A non Linear System:
ẋ = f(x) + g(x)u

rás
Where f : D → Rn , and g : D → Rnxp , are suciently smooth on a domain
is said to be Feedback Linearizable (or input-state
Rn ,

Bor
D⊂
linearizable, if there exists a dieomorphism T : D → Rn such that
Dz = T(D) contains the origin and the change of variables z = T (x)
transform the equation 1 into the form:
los
ż = Az + Bγ(x)(u − α(x))
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 25 / 72
 Feedback Linearization
Now, we would like to solve the tracking problem for the previous system:

7
201
x˙1 = a · sin(x2 )
x˙2 = −x12 + u
y = x2

rás
If we still use the state feedback linearization, we should obtain:

Bor
z1 = x1 ⇒ z˙1 = z2
z2 = a · sin(x2 ) = x˙1 ⇒ z˙1 = v
los
1 z2
u = x12 + v ⇒ y = sin−1 ( )
acos(x2 ) a
Car

NOT SO GOOD!, because Linearizing the state equation doesn't

necessarily linearize the output equation. However if WE set up u = x12 + v
GRAPHS/UIS.p
we got:
Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 26 / 72
 Feedback Linearization

7
z1 = x1 ⇒ z˙1 = z2

201
z2 = a · sin(x2 ) = x˙1 ⇒ z˙1 = v
1 z2
u = x12 + v ⇒ y = sin−1 ( )

rás
acos(x2 ) a
NOT SO GOOD!, because Linearizing the state equation doesn't

Bor
necessarily linearize the output equation.
However if WE set up u = x21 + v we got:
los
x˙2 = v
Car

y = x2
which is OK, but, x1 is unobservable from y . WE have to check that x1 is
well behaved. x˙1 = a · sin(x2 ), check the dynamics. if y = yd = cteGRAPHS/UIS.p
,
therefore: x1 (t) = x1 (0) + t · a · sin(yd ), It is unbounded.
Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 27 / 72
 Input Output Linearization
Consider the single input, single output non linear dynamic system

7
201
ẋ = f (x) + g (x)u
y = h(x)

rás
Where: f , g , and h are suciently smooth in a Domain D ⊂ Rn . The
mappings f(x) : D → Rn , and g(x) : D → Rn , are called Non Linear

Bor
vector eld on D .
     
x˙1 f1 (x1 , x2 , ..., xn ) g1 (x1 , x2 , ..., xn )
los
x˙2  f2 (x1 , x2 , ..., xn ) g2 (x1 , x2 , ..., xn )
     
Car

.  .   . 
 = + u
.  .   . 
     
.  .   . 
x˙n fn (x1 , x2 , ..., xn ) gn (x1 , x2 , ..., xn ) GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 28 / 72
 Input Output Linearization- SISO

Consider the single input, single output non linear dynamic system

7
201
ẋ = f (x) + g (x)u
y = h(x)

rás
Where: f, g, and h are suciently smooth in a Domain D ⊂ Rn . The

Bor
mappings f : D → Rn , and g : D → Rn , are called vector eld on D.
The rst output derivative ẏ is given by:
los
∂h
[f (x) + g (x)u] = Lf h(x) + Lg h(x)u
def
ẏ =
∂x
Car

Where:
∂h ∂h
Lf h(x) = f (x) Lg h(x) = g (x) GRAPHS/UIS.p
∂x ∂x
Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 29 / 72
 Lie Derivative of h with respect to f or along f

7
201
Lie Derivative: NonLinear vector elds f (x) and g (x)
Lie Derivative of h with respect to f or Along f

rás
∂h
Lf h(x) = f (x)

Bor
∂x
Lie Derivative of h with respect to g or Along g
los
∂h
Lg h(x) = g (x)
∂x
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 30 / 72
 Lie Derivative Concept : Mathematical Tool

7
Concept of Lie Derivative

201
Assume that h(x) is a escalar function; Like Energy h(x)
How is the function vary on time?. How the Energy vary on time?

rás
dh(x) ∂ h dx
= ·
dt ∂ x dt

Bor
 
dh(x) ∂ h dx ∂h ∂h ∂h
= · = , , ...., (f (x))
dt ∂ x dt ∂ x1 ∂ x2 ∂ xn
los
 
∂h ∂h ∂h
Lf h = , , ...., (f (x))
Car

∂ x1 ∂ x2 ∂ xn

Where: f (x) Non Linear vector eld f (x)

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 31 / 72
 Lie Derivative Concept : Mathematical Tool

The rst output derivative: ẏ is given by:

7
201
∂h
[f (x) + g (x)u] = Lf h(x) + Lg h(x)u
def
ẏ =
∂x
if Lg h(x)u = 0, then ẏ = Lf h(x), its independent of u

rás
The second derivative: ÿ

Bor
∂ (Lf h)
[f (x) + g (x)u] = L2f h(x) + Lg Lf h(x)u
def
ÿ =
∂x
los
if Lg Lf h(x)u = 0, then ÿ = L2f h(x), its independent of u
...
The third derivative: y ..... Repeating the process, if:
Car

1
f h(x)u = 0,
Lg Li− i = 1, 2, 3, ..., k − 1
1
Lg Lk−
f h(x) 6= 0 GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 32 / 72
 Lie Derivative Input Output Linearization

7
201
...
The u does not appear in: y , ẏ , ÿ , y ,...,y k−1 , until :
1

rás
y (k) = Lkf h(x) + Lg Lk−
f h(x)u

Where: Lg Lk− 1
h(x)u 6= 0, u appears explicity.

Bor
f
The system is input - output linearizable and reduced to a chain
of k integrators:
los
y (k) = v , ⇒ (Chain − of − K − Integrators)
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 33 / 72
 Lie Derivative Concept: Mathematical Tool

7
201
Consider the single input, single output non linear dynamic system

rás
ẋ = f (x) + g (x)u ∀ x(t) ∈ D, (D ⊂ Rn )
y = h(x)

Bor
Can be transform into full linearized model or at least partially linear model.
Order of the system: Is the Number of time we have to dierential the
los
Output y until the input u appears in the output (Explicity appears in the
output). WE called as Relative degreee of the output y
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 34 / 72
 Lie Derivative Concept: Mathematical Tool
Example: Consider the non linear dynamic system:

7
201
x˙1 = x2
x˙2 = −x1 + ε(1 − x12 )x2 + u

rás
y = x1

What is the relative degree of the system?. Where u, Explicity appears in

Bor
the output. los
dy
= x˙1 = x2
dt
Car

d 2y
= x˙2 = −x1 + ε(1 − x12 )x2 + u
dt 2
The Relative degree is TWO (2) GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 35 / 72
 Order "n" system: Assume relative degree is k ≤ n

7
Applying Lie Derivative:

201
dy (t) ∂ h dx ∂h
= · = [f (x) + g (x)u]
dt ∂ x dt ∂x

rás
dy (t) ∂ h dx ∂h ∂h
= · = f (x) + g (x)u
dt ∂ x dt ∂x ∂x

Bor
dy (t) ∂ h dx
= · = Lf h + (Lg h) · u
dt ∂ x dt
Assuming that (Lg h) = 0, it doesn't appear u, Not Yet.
los
Again, applying Lie Derivative:
Car

d 2 y (t) d
= (Lf h) = L2f h + (Lg Lf h)u
dt 2 dt
GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 36 / 72
 Order "n" system: Assume relative degree is k ≤ n

7
Again....Applying Lie Derivative: until u appears EXPLICITY.

201
d 2 y (t) d
= (Lf h) = L2f h + (Lg Lf h)u
dt 2

rás
dt
d 3 y (t)
= L3f h + (Lg L2f h)u
dt 3

Bor
..
.
d k y (t) 1
= Lkf h(x) + (Lg Lk−
los
f h(x))u
dt k
Until u , appears Explicity.
Car

If k = n, it is full linearized system, if not, k < n, it is parcially

linearized
GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 37 / 72
 The state feedback control

7
201
Until u appears EXPLICITY. Therefore, the state feedback control u
1

rás
u= 1
[−Lkf h(x) + V ]
Lg Lk−
f h(x)
Input-output map to:

Bor
d k y (t)
=V
dt k
it is a chain of k integrators. The integer k is called the Relative
los
degree of the system.
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 38 / 72
 Lie Bracket

7
201
Let f and g two vector eld on Rn . Lie Bracket of f and g is a third
vector eld, dened by: [f,g]

rás
∂g ∂f

Bor
[f, g] = ·f− ·g
∂x ∂x
[f, g] = 5g · f − 5f · g
[f, g] = adf g
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 39 / 72
 Example of Lie Bracket

Let f and g two vector eld on Rn . Lie Bracket of f and g is a third

7
vector eld, dened by: [f,g]

201
Consider the system:

rás
ẋ = f (x) + g (x)u

Bor
with the two vector eld f (x) and g(x) dened by:
−2x1 + ax2 + sin(x1 ) 0
   
f = g=
cos(2x1 )
los
−x2 cos(x1 )
The Lie Bracket can be computed as:
Car

[f, g] = 5g · f − 5f · g
GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 40 / 72
 Example of Lie Bracket

Lie Bracket of f and g is a third vector eld, dened by: [f,g]

7
201
with the two vector eld f (x) and g(x) dened by:
−2x1 + ax2 + sin(x1 ) 0
   

rás
f = g=
−x2 cos(x1 ) cos(2x1 )
The Lie Bracket can be computed as:

Bor
[f, g] = 5g · f − 5f · g
los
0 0 −2x1 + ax2 + sin(x1 ) −2 + cos(x1 ) 0
     
a
= −
−2sin(2x1 ) 0 −x2 cos(x1 ) x2 sin(x1 ) −cos(x1 ) cos(2x1 )
Car

acos(2x1 )
 
[f, g] =
cos(x1 )cos(2x1 ) − 2sin(2x1 )(−2x1 + ax2 + sin(x1 ))
GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 41 / 72
 Properties of Lie Bracket

Bilinearity:

7
201
[α1 f1 + α2 f2 , g ] = α1 [f1 , g ] + α2 [f2 , g ]
[f , α1 g1 + α2 g2 ] = α1 [f , g1 ] + α2 [f , g2 ]

rás
where: f,f1 ,f2 ,g,g1 and g2 are smooth vector elds, and α1 and
α2 , are constant scalars.

Bor
Skew-Commutativity:
fg
[ , ] = −[ , ] gf
los
Jacobi Identity:
Car

Ladf g h = Lf Lg h − Lg Lf h
Where h(x) is a smooth scalar function of x GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 42 / 72
 Example: Van der Pol System

7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 43 / 72
 Example

7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 44 / 72
 Example

7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 45 / 72
 Example

7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 46 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 47 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 48 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 49 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 50 / 72
 Full State Linearization

7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 51 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 52 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 53 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 54 / 72
 Example 2

7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 55 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 56 / 72
 Example 3

7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 57 / 72
 Theorem 13.2

7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 58 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 59 / 72
 Example: Single Link Manipulator

7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 60 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 61 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 62 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 63 / 72
 Example: DC Motor

7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 64 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 65 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 66 / 72
 State Feedback Control

7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 67 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 68 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 69 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 70 / 72
 Tracking

7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 71 / 72
 7
201
rás
Bor
los
Car

GRAPHS/UIS.p

Carlos Borrás P. Ph.D., MSc. (EIM-UIS) Nonlinear Control Engineering I October 22, 2020 72 / 72