Feedback Principles
+
-1
0 5 10 + -1
0 5 10
Linear system
1 0.5 and frequenc
with input am
u2 0 y2 0
2X input ⇒
-1
0 5 10
-0.5
0 5 10 Allows us t
2 2
percen-tag
u1 + u2
y1 + y2
response.
0 0 of input am
-2 -2
Limitation:
0 5 10 0 5 10
holds up to
Figure 1: A demonstraion of the superposition theorem.
M. Farooq Haydar
18 Oct 04
(IST) Control Systems
R. M. Murray, Caltech CDS
March 29, 2018 4 / 35
Mathematical Models: LTI Systems
The LHS of (1) specifies the homogeneous solution (free response), which
is the solution to
The LHS of (1) specifies the homogeneous solution (free response), which
is the solution to
Characteristic Equation:
The solution to the characteristic polynomial
a(s) = sn + a1 sn−1 + a2 sn−2 + ... + an−1 s + an , (3)
is called a pole of the system and describes the homogeneous solution.
Xn
y(t) = Ck esk t ,
k=1
Xn
y(t) = Ck esk t ,
k=1
Xn
y(t) = Ck esk t ,
k=1
Xm
y(t) = Ck (t)esk t ,
k=1
where
I Ck (t) is a polynomial in t and k < m.
y(t)
y(t)
0.5 0.5
1
0 0 0
0 0.5 1 0 2 4 0 0.5 1
Time t Time t Time t
s=0 s = −1 s=1
1 20
1 y(t)
y(t)
y(t)
0 0 0
−1
−1 −20
0 5 10 15 0 5 10 15 0 5 10 15
Time t Time t Time t
s=i s = −0.2 + i s = 0.2 + i
Figure 2.1: Examples of exponential signals. The top row corresponds to exponential signals
with aFigure
real exponent,
2: Time and responses
the bottom rowof corresponds
the roots toofthose with complexequation.
characteristic exponents. The
dashed line in the last two cases denotes the bounding envelope for the oscillatory signals.
In each case, if the real part of the exponent is negative then the signal decays, while if the
M. Farooq Haydar (IST) Control Systems March 29, 2018 11 / 35
Mathematical Models: Transfer Functions
Lets now consider again (1)
y n + a1 y n−1 + a2 y n−2 + ... + an y = b0 um + b1 um−1 + ... + bm u. (4)
If a(x) 6= 0
b(s) st
y(t) = y0 est = e = G(s)est = G(s)u(t).
a(s)
M. Farooq Haydar (IST) Control Systems March 29, 2018 12 / 35
Transfer Function
The transfer function of an LTI system
y(t) = G(s)u(t).
y(t) = G(s)u(t).
Remember:
A transfer function is only defined for an LTI system!
*0
sY (s) −
Y (0)
= U (s),
Y (s) 1
= ,
U (s) s
1
=⇒ G(s) = .
s
Y (s)
= s,
U (s)
=⇒ G(s) = s.
Xm
y(t) = Ck (t)esk t + G(s)est .
k=1
0.2
0.1
Output
−0.1
0 5 10 15
0.5
Input
−0.5
−1
0 5 10 15
Time
Figure 4: Two solutions, with same input. (Particular/forced solution is the same,
Figure 2.2: Two responses of a linear time-invariant system to a sinusoidal input. The dashed
while homogeneous
line shows the outputsolution
when theisinitial
different due toarethechosen
conditions initialsoconditions.)
that the output is purely
sinusoidal. The full lines show the response response for the initial conditions y(0) = 0 and
y′ (0) = 0. The transfer function G(s) = 1/(s + 1)2 .
M. Farooq Haydar (IST) Control Systems March 29, 2018 18 / 35
Review: Fourier Series
Figure 6: Convolution of an input signal x(t) with the impulse response h(t).
Figure 7: Calculation of output from input and the impulse response, both in time
and complex frequency (s) domain. (∗ is the convolution operation.)
M. Farooq Haydar (IST) Control Systems March 29, 2018 23 / 35
Transfer Functions & BlockFEEDBACK
2.2. USING Diagrams TO IMPROVE DISTU
G
u y u
G1 G2
G
y1 = G1 u
y2 = G2 y1
=⇒ y = G2 G1 u
G1 u
Σ
y u y
G2 Σ
G2
u e y
Σ G1
y
Σ
−G2
r e u y
Σ C(s) Σ P(s)
−1
re 2.4: Block diagram of simple feedback system. The controller transfer functio
and the process transfer function is P(s). The process output is y, the external sig
he reference r and the loadYdisturbance
(s) = P (s) v.
(V (s) − C(s)Y (s))
P (s)
Y (s) = V (s)
1 + P (s)C(s)
ature, blood pressure and other Pimportant (s) variables constant. For ex
Gyv (s) =
pillary reflex guarantees that1 the + P light
(s)C(s)intensity of the retina is reas
nt in spite of large variations inControl
M. Farooq Haydar (IST)
the ambient
Systems
light intensity. Keeping
March 29, 2018
va
27 / 35
Feedback
2-10 for Disturbance Attenuation
CHAPTER 2. FEEDBACK PRINCIPLES
2 2
kp
ωc
1 1
y
y
0 0
0 2 4 6 8 0 2 4 6 8
0
0
kp ωc
−0.5 −0.5
u
u
−1 −1
0 2 4 6 8 0 2 4 6 8
Normalized time, at Normalized time, at
(a) Proportional control (b) Proportional-Integral (PI) control
Figure 2.5: Step responses for a first-order, closed loop system with proportional control
Figure 8: Disturbance Rejection with Proportional and PI control. (PI control is
(a) and PI control (b). The process transfer function is P = 2/(s + 1). The controller gains
designed
for proportional control are k p =ξ0,=0.5,
with damping-ratio 0.707 and
1 and undamped
2. The natural
PI controller frequency
is designed using equa-
ωction
= (2.19)
0.707,with
1, ζand 2.)
c = 0.707 and ωc = 0.707, 1 and 2, which gives the controller parameters
M. Farooq Haydar (IST) Control Systems March 29, 2018 28 / 35
CHAPTER 2. FEEDBACK PRIN
Feedback for Reference/Command Tracking
v
r e u y
Σ C(s) Σ P(s)
−1
re 2.4: Block diagram of simple feedback system. The controller transfer functio
and the process transfer function is P(s). The process output is y, the external sig
he reference r and the loadYdisturbance v. (R(s) − Y (s))
(s) = P (s)C(s)
P (s)C(s)
Y (s) = R(s)
1 + P (s)C(s)
ature, blood pressure and other important variables constant. For ex
P (s)C(s)
Gyr (s) =
pillary reflex guarantees that1 the + P light
(s)C(s)intensity of the retina is reas
nt in spite of large variations inControl
M. Farooq Haydar (IST)
the ambient
Systems
light intensity. Keeping
March 29, 2018
va
29 / 35
Feedback for Reference/Command Tracking
2.3. USING FEEDBACK TO FOLLOW COMMAND SIGNALS 2-13
1.5 1.5
1 1
y
y
ζc
0.5 ωc 0.5
0 0
0 2 4 6 0 2 4 6
6 3
ζc
4 2
ωc
u
u
2 1
0 0
0 2 4 6 0 2 4 6
Normalized time, ωc t Normalized time, ωc t
Figure 2.6: Responses to a unit step change in the command signal for different values of
Figure 9: Reference Tracking with PI control.
the design parameters ωc and ζc . The left figure shows responses for fixed ζc = 0.707 and
ωc = 1, 2 and 5. The right figure shows responses for ωc = 2 and ζc = 0.5, 0.707, and 1. The
process parameters are a = b = 1. The initial value of the control signal is k p .
M. Farooq Haydar (IST) Control Systems March 29, 2018 30 / 35
Two Degrees of Freedom Control
6. FEEDBACK AND FEEDFORWARD 2-25
Controller
Fu v
uff
r ym e ufb u y
Fm Σ C
− Σ Σ P
−1
Figure 2.18: Block diagram of a closed loop system where the controller has an architecture
Figure
with two degrees 10: TwoThe
of freedom. degrees
desiredofresponse
freedomymcontrol
and thearchitecture.
feedforward signal uff are
generated from the reference r. The feedback controller C acts on the control error e = ym −y
and generates the feedback control signal ufb .
Controller v
−Fv
r u y
Σ C Σ P1 Σ P2
−1
Figure 2.19: Block diagram of a system with a control architecture that combines feedback
Figure 11: Two degrees of freedom control architecture, where disturbance can be
with feedforward from a disturbance v that can be measured.
measured and a correction can be applied.
s
Σ
Primary Output
− G(s)
microphone
Secondary n
Filter
microphone
(a) Schematic diagram
Figure 12: Noise cancellation using feedforward.
Figure 2.20: Schematic and block diagrams fo
Noise Cancellation
Noise cancellation is a common example of the us
M. Farooq Haydar (IST)fects of disturbances.
Control Systems Consider, for example,
March 29, 2018 33a/ pil
35
Magnitude (
-50
Case-100Study
Example:
-150 Satellite Attitude Control
10 -1 10 0 10 1 10 2 10 3
Frequency (rad/s)
Closed loop step response
1
Amplitude
0.5
0
0 0.5 1 1.5 2 2.5
Time (seconds)
Control effort for step response
40
Amplitude
20
0
0 0.5 1 1.5 2 2.5
Time (seconds)
Magnitude (
Case Study 10
-10 closed loop
closed loop + prefilter
open loop
Example: Inverted
10 Pendulum -20
-3 -2 -1 0 1 2 3 4
10 10 10 10 10 10 10 10
Frequency (rad/s)
1
Amplitude
0.5
-0.5
0 2 4 6 8 10 12 14 16
Time (seconds)
0.4
Amplitude
0.2
-0.2
0 1 2 3 4 5 6 7
Time (seconds)
Figure 9: The closed-loop bode response, the unit-step response for the out-
Figure
put, 14:unit
and the Step
stepresponse
response forwith twoofdifferent
the input controllers.
the inner loop design in
HW B.18.
M. Farooq Haydar (IST) Control Systems March 29, 2018 35 / 35