Feedback control

·
negative feedback loop reacts to disturbances so as to decrease the

effectthey have

o
eg. 1. quadcopter vice

2. regulating blood glucose levels

·
Positive Feedback loop reacts to disturbances to as to increase the

effect they have. It often causes instability.

e.g. microphone at distanced from loudspeaker. Loudspeaker plays

amplifed vernor sound
a at mic.

amps

· example:making a robot more

from point A to B (Feedforward vs Feedback)
FeedForward control
-

he-melt
- >

A plan:given how to choose

r, a to give the desired
output?
requires very good knowledge of system (environment

A Issues:friction, something unexpected?

 -
Feedback control

/Fe
A controller:given measurement position
of y, compute enore,

update 4 to
try to reduce error

A Issues:norsy measurements, delays

-
Feedback vs.
Feedforward

Feedback control Feedforward control

closed
*

loop A
open loop

Acts
A on deviation Acts
* on plans

Robust to model
* uncertainty a sensitive to model uncertainty
Risk ofinstability
* no
* risk for instability
sensitue
* to measurement Insensitive to
* measurement

noise house

Components of a control system

·
systems we to
want control

-Plant:systems be
to controlled

-control
Input:signals we can
modify
-ourput:signals we want to control

Disturbance:Inputs we cannot modify

·
Actuators and sensors
 -
Actuator:physical system that implements control input

-sensor:physical system thatmeasures plant output

·
Feedback control loop

-controller:system that we design

*Inputs:
-> measurement of plant(from sensor (

-> Reference input specifying desired behaviour

A output:control signal (input to plant)

-

eg:
1
heating a room

opening a door
or window

when it furn air in

nearer
heater on
/OFF the room
dial of the
temperature
thermostat

I
temp
sensor

quantisation
air may nothave uniform
-

temp,
systematic bias From location ofsensor
in the room

Basic arms
of control.

· stabilisation

-many systems are unstable:undesireable in

engineering systems
 ~unstable:small changes to the
inputs or disturbances can cause

large differences in
system behaviour

-
Am:deagn a controller to stabolise an unstable system.

Regulation
·

-Aim: to maintain system outputs near a constant value despite

dynamics and disturbances.

Eg:Cruise control - maintain constant

speed
·

typical disturbances; road, surface of road, wind

slope of
·
Reference tracking
-given:a class of constant
reference equal, eg. piecewise agnals

or piecewise slowly varying signals.

-

Alm: design controller so

system output tracks
every

reference equal in the

given class (within
acceptable error

disturbances.
In
presence of

regulation problem (ref.class:constant signals)

-special case:

- and
Transient steady-state response

m y
the transient
resp.
is

Reference tracking:errorlt) reference(t) =

-

ourpht (t)
*
Transient Response:how does error behave soon after a

sudden change in reference?

-> how long does it take to settle down?

- how much does fluctuate?

it

steady. State
*
Response:How small is emort for larget?
Transfer Functions in a feedback
loop
simplied Feedback system model
·

Distrubance horse

controller

systems:

*All systems modelled by linear

constant coefficient differential

equations ((((DE)

plant
*
transfer Function:PIs)
controller transfer Funchon:(es)
A

A
Feedforward transfer funcion:FCS) Loften ECsI=1)

Input signals: output signals:

-

Areference w *
measurement y

disturbance
* v *Process output in retal

A measurement noise W A process input V (nu)

control
*
synal U

A
er ror t v
=
-

E r
= -
x

-
Arms of control

stabilization
* Disturbance
* rejecten

Robustness to model
Improving
* transcent response A
uncertainty
reference
*
tracking
Equivalent Transfer Funchous
·

-point of
view of angle output:(via linearity (

YCS) Gny(S) Rcss

=

+
Gry(s) Viss +
GwyCS) WisI
-

point of
view angle
of input output pair:

Yess GryC)
=

RCSS

Approach:(via linearity
-

set all
A to except the of
inputs zero input interest

*End transfer Function from

that interest
inputto outputof
example
-

1.
Transfer Funchon From reference to output

Myin

YCs) PCs).=
Ucs1 PCSI. CCs). ECs)
=

=PCs) Cass [FIsIRCss -YCss]

Grycs)
I"s
=

2.
Transfer Funchon from disturbance
in error e

0 O
=

IIIII),
0
0
"I
=

L
mult by -1

Ecs=-Yess = -
Pass [Hass+Wass]
--Pass [CascEcsit Vass]
Eiss+ PasiCiscEcss=-PasiUcsI

Grecss-Ecs
 3. Transfer Funchon From distrubance
v to error e

general form:

:
IM I
mutty See I

crestsin
Tee
~

-Pas
Performance specifications
· How does a test
aqual at an interest
inputof affect a
corresponding output
of interest?

Key testagnals:Unit steps, sinusoids (at different frequencies)

·
Step response:output system
of when input is unit step

withthe E

Parameters:
*

Risehme (Tr) steady-state

* value:

Peak
* hme (Tp) 1a
carnal:
((t)

Settling
* time (Ts) &
Percentage overshoot:

e
Mp a
=
 Response
·
sinusoidal
to
input

If real III system has a

transper Funchon of Ecs) and

inputis cos(wt 1), then output

+
is (Cia)(cos(wt+ 4+ (Eciwl)
↑ ↑
gain phase shipt

Frequency Response:Elitol
-

is evaluation transfer Furchon

of of purely imaginary inputs.
Freg, response us how
tells the
system response to every sinusoidal
-

test agral.
Typical
·

Frequency Response Behaviour

-
observations;

Disturbances
*

usually slow varying (low-frequency)

measurement
* noise
usually high-frequency

-motrates
frequency response-based performance specification
-

Eg. constraints on
Free, resp. equivalent
of system
From
* disturbance to
output:
IEryciw) I should be small
for small w.

From
* noise
measurement to
output:
1 Eru(iw) I should be small large w.
for

· Abstract Design Problem

-generalized plant:abstraction of Exed parts design

of problem
-

Inputs:
-

outputs:
vector
* input w vector
* output z 6F

of all exogeneous all signals we want

signals small (error. Cont

* control input U