01-Introduction.txt

Lecture: Introduction
Automatic Control 1
Prof. Alberto Bemporad
University of Trento
Academic year 2010-2011
Faculty of Engineering
Prof. A. Bemporad (University of Trento)
Academic Year 2010/2011
Outline
Lecture outline
What is automatic control ?
Application examples
Course information
2 /50
What is control ?
What is control ? A real life example ...
water knobs
desired temperature
r(t)
disturbance
shower
d(t)
water inflow
water temperature
u(t)
y(t)
skin sensations
Water inflow u(t) must be controlled to reach and maintain
the desired temperature r(t)
Sensors on skin measure water temperature y(t)
Water inflow u(t) manipulated so that y(t)≈r(t) …
… in spite of flow and temperature fluctuations d(t)
3 /50
steering wheel
desired lateral
displacement r(t)
vehicle on road
steering angle
lateral displacement
eyes watching current displacement
Steering wheel must be controlled to reach and maintain
the desired lateral displacement r(t) within the lane
(e.g.: staying in the middle of the lane)
Eyes measure current later displacement y(t)
Steering wheel u(t) manipulated so that y(t)≈r(t) …
… in spite of changes of road curvature and of r(t)
4 /50
embedded control unit
reference
process
manipulated input
measurements
How to control the inputs u(t) to the process automatically
to make the output y(t) track the given reference r(t) ?
performance
How to exploit the measurements of y(t) to track the reference
r(t) in spite of disturbances d(t) acting on the process ?
robustness
5 /50
What is control ? A real life example …
Mom is washing hands,
she drains hot water
(=disturbance entering
the process)
6 /50
Examples of control systems
Application areas of control engineering
Aeronautics & aerospace
Automotive
Manufacturing
Process control (chemical, pharmaceutical,
steel, pulp & paper, ...)
Power electronics
Telecommunications
Environmental systems
Financial engineering
Supply chains
Power networks
7 /50
Adaptive Cruise Control (ACC)
8 /50
Electronic Control Unit
road slope
desired speed
engine torque/brakes
vehicle
speed/position
9 /50
Lane Warning
/DQH 'HWHFWLRQ +: SODWIRUP
HALF - HAptical Lane Feedback
5LFHUFKH
Optical sensor detects and tracks lane markings
VIP ECU
9LVLRQ driver by: components
9LVLRQ 6\VWHP
• Automotive
Incipient lane change signalled to the 6\VWHP
A N E WA RN I N G
&OXE RI
([SHUWV
• Power supply: 12Vbat - 500mA
ptical sensor detects and tracks lane markings.
• rumble strip)
- Haptic signal on driver seat (simulation of analogue video inputs
- driver feedback on steering wheel
gnals theHapticincipient change of lane&RQWURO
by means of:• DSP @ 60Mhz
- Acoustic directional signals 6\VWHP
Haptic feedback on steering wheel (HALF)
• CAN + RS-232 interface
Haptic signal on driver seat (simulation of rumble strip)
• many configurable Digital I/ Os
Acoustic directional automatically maintained
signals
Lane can be
• 2MB DRAM + SRAM
6WHHULQJ 6\VWHP
Orbassano
17th Dec. 2002
electric power steering
Automatic Control 1 & Process Engineering
10 /50
11 /50
displacement
Driver
Torque
Demand
+ $ Brava 1600)
Vehicle (Fiat / )
Steer
actuated
lateral
+$/) GHPRQVWUDWRU KDV y(t) GHYHORSHG E\ &5)
+$/) GHPRQVWUDWRU KDV EHHQ GHYHORSHG E\ &5)
EHHQ
,QJ 9 0XUGRFFR
12 /50
Idle speed control
Objective:
maintain the engine speed at a given rpm
Power steering + air conditioning
Torque disturbance: power steering
13 /50
Active suspensions
Active Suspension System
Ford Mercur XR 40i
active
suspensions
passive
14 /50
(Semi)active suspensions
Ms=suspended	
  mass
Mus=unsprung	
  mass
x4	
  sprung	
  mass
velocity
suspension
deflection
x2	
  unsprung	
  mass
tire	
  
active	
  suspensions
semiactive	
  suspensions
15 /50
Traction control
Problem:	
  Improve	
  driver's	
  ability	
  to	
  control	
  a	
  vehicle	
  under	
  adverse	
  
external	
  conditions	
  (wet	
  or	
  icy	
  roads)
indoor	
  tests	
  at	
  ice	
  arena	
  (μ¼	
  0.2)
2000	
  Ford	
  Focus,	
  2.0l	
  4-­‐cyl	
  engine
5-­‐speed	
  manual	
  transmission
16 /50
Air to Fuel Ratio (AFR) control
Control objectives:
Maintain the stoichiometric Air to Fuel Ratio (AFR) and avoid oxygen saturation
(empty or full catalyst)
Fuel
Tank
Fuel injector
(Actuator)
Air/Fuel
mixture
Exhaust gas
Delay
UEGO
(Sensor)
Three Way Catalyst
17 /50
Control of suspensions
(Courtesy	
  of	
  Daimler-­‐Chrysler)
18 /50
Active steering
Courtesy of
19 /50
Segway Human Transporter
www.segway.com
20 /50
The Segway™ Human Transporter (HT) is the first
self-balancing, electric-powered transportation device.
With dimensions no larger than the average adult body
and the ability to emulate human balance, the Segway
HT uses the same space as a pedestrian, and can go
wherever a person can walk.
Dynamic stabilization is the essence of the Segway
Human Transporter (HT). Dynamic stabilization
enables Segway HT to work seamlessly with the body’s
movements.
Gyroscopes and tilt sensors in Segway HT monitor a user’s center of
gravity at about 100 times a second. When a person leans slightly
forward, Segway HT moves forward. When leaning back, Segway HT
moves back.
21 /50
22 /50
Short historical notes
Water level controllers date back to ancient Greece
(Ktesibios of Alexandria, water clock, 270 B.C.), (Philon of Byzantium, 250 B.C.)
(Heron of Alexandria, ≈100 A.D.)
Other examples in the 17th-18th century
(Cornelis Drebbel’s incubator, 1620), (Edmund Lee’s self-regulating wind machine, 1745)
(Thomas Mead’s lift tenter and speed controller, 1787)
James Watt’s flyball governor (1788)
Used to regulate the speed of steam engines.
If speed increases, flyballs spread apart and
the steam flow through the throttle is reduced.
And vice versa.
Frequency domain (1930-1950)
(H. Nyquist, H.W. Bode, N.B. Nichols)
State-space and optimal control (1950-1980)
(R. Kalman, R. Bellman)
Nonlinear, robust, adaptive, optimization-based, (…) control (1980-today)
23 /50
Is control needed after all ?
Friday November 7, 1997
Baby Benz falls foul of a moose
“Mercedes is recalling 3,000 brand-new
A Class mini cars to correct stability
problems revealed in Scandinavian
tests.”
A Class failed the “moose test”, simulating a sudden swerve
to avoid a moose on the road.
The recall of about 3,000 cars followed motoring press
reports which indicated the A Class is unstable, and likely
to roll over, in extreme lane-changing tests.
Problem solved by introducing the Electronic Stability
Program (ESP), which automatically manipulates brakes
(and engine torque) to keep skid under control.
24 /50
EC – DG INFSO – SMART 2007/047
Monitoring and control: today's market, its evolution till 2020 and the impact of ICT on these
WORKSHOP: Brussels the 9th of October 2
Is control needed Worldwide?Monitoring & Control Markets
after all
WORKSHOP: Brussels the 9th of October 2008
Worldwide Monitoringby application
& Control Markets
Total world: !187,9 bn
by application
Factory automation, the sum of Manufacturing plus
Worldwide Monitoring & Control market market with 58
Process industries, remains the main
industries
Total
billion euros, comparable to
billion):
(=188 billions euro)– sum of ManufacturingVehicles (56dominates with 2/3 of world: !187,9 bn
Factory automation, the With 38 billion euros out of 58, Services
56,4
Vehicles
total Factory market market.
Process industries, remains the mainAutomation with 58
– to Among Services the sub segment Integration installation and
Integration,
billion euros, comparable Vehicles (56 billion):
for 50% of
Monitoring & control oftraining accountsdominatesthem. 2/3 of
≈ semiconductor industry
Industry
– With 38 billion euros out 58, Services
with
total Factory Automation market.
≈ 2 x mobile phoneTogether, three application markets, Vehicles,
industry
Among Services the sub segment Integration installation and
Manufacturing and Process industries represent 60% of
f t i
i d ti
t Process
training accounts for 50% of them.
total Monitoring & Control market.
Still growing, young areas with great potentials
Healthcare
Together, three application between 10 and 20 billion
Ranked
(power fgrids, and Process markets, Vehicles,order : euros the next three
building industries representtenvironment)
automation,
Critical
Manufacturing d P applications tmarkets are in 60% of
Logistics &
transport
Ranked between 10 and 20 billion euros the next three
Environment
applications markets are in one, Home is, for the moment, a small niche market.
Last order :
Critical infrastructures
Logistic & transport
A report &
Logistic & transport a presentation prepared by:
26,3
18,0
10,9
Power grids
Building
Homes
18,3
57,8
Hard
Serv
Available for download: http://www.decision.eu/smart2007.htm
Last one, Home is, for the moment, a small niche market.
Household
appliances
A report & a presentation prepared by:
31,5
Infrastructures
Process
– Healthcare
25 /50
Process control
Control is heavily used in the process industries
wi ! j
the controller does its best to track rj, sacrificing ri tracking if
necessary. If w j = 0 , on the other hand, the controller completely ignores
deviations rj–yj.
Usually slow processes (heat transfer, chemical reactions, distillation, etc.)
Choosing the weights is a critical step. You will usually need to tune your
controller, varying the weights to achieve the desired behavior.
Usually several input and output variables
As an example, consider Figure 1-3, which depicts a type of chemical reactor (a
CSTR). Feed enters continuously with reactant concentration CAi. A reaction
takes place inside the vessel at temperature T. Product exits continuously, and
contains residual reactant at concentration CA (<CAi).
The reaction liberates heat. A coolant having temperature Tc flows through
Example: Continuous stirred-tank reactor
coils immersed in the reactor to remove excess heat.
Figure 1-3: CSTR Schematic
Manipulated inputs:
reactant concentration CAi in feed
From the Model Predictive Control Toolbox point for view, T and CA would be
coolant CAi and Tc would
plant outputs, and temperature Tcbe inputs. More specifically, CAi would be
an independent disturbance input, and Tc would be a manipulated variable
(actuator).
Controlled outputs:
There is one manipulated variable (the coolant temperature), so it’s impossible
reactant concentration CAi in vessel
to hold both T and CA at setpoints. Controlling T would usually be a high
vessel temperature T
priority. Thus, you might set the output weight for T much larger than that for
CA. In fact, you might set the CA weight to zero, allowing CA to float within an
Prof. A. operating region (to of Trento)
acceptableBemporad (Universitybe defined by constraints).
www.abacusengg.in
26 /50
Control systems example: Ball and plate
27 /50
commands
station
ball &
plate
28 /50
Robotics
29 /50
Snake Robot (Biorobotics and Biomechanics Lab, Technion, Israel)
30 /50
Functional electrical stimulation
31 /50
Aeronautics and aerospace
Control systems heavily used in:
Aircrafts (roll, pitch, yaw)
Helicopters
Satellites
32 /50
Aeronautics and aerospace. Quadcopter example
Parrot AR Drone
• 6 DoF Inertial Measurement Unit
(3 accelerometers, 3 gyros)
•468MHz ARM9 CPU running Linux
•Ultrasound altimeter/ground detector
•Two cameras:
- Vision-based stabilization (176x144)
- Wifi streaming (640x480)
7 (27)
Prepared
Title
M. Lefébure
AR Drone Developer Guide
Approved
Date
Revisio
File
AR Drone Developer Guide Release 1.0.doc
quadcopter
05/01/2010
flight control unit
desired angles
engine torques
angles/positions
2D-tag: magnetic cylindrical tag
33 /50
• Manipulated inputs
– total force u
– torques τθ , τφ, τψ
• Outputs to regulate
−u sin θ − β x
u cos θ sin φ − β y
34 /50
Outputs: position (x,y,z)
Inputs: torques and force
35 /50
36 /50
Financial engineering. Example: Option hedging
• The	
  financial	
  institution	
  sells	
  a	
  synthetic	
  option	
  to	
  a	
  customer	
  and	
  gets	
  y(0) (€)
• Such	
  money	
  is	
  	
  used	
  to	
  create	
  a	
  portfolio	
  y(t) of	
  underlying	
  assets	
  (e.g.	
  stocks)
whose	
  quantities	
  at	
  time	
  t	
  are	
  u1(t),	
  u2(t),	
  ...,	
  un(t)
• At	
  the	
  expiration	
  date	
  T,	
  the	
  option	
  is	
  worth	
  the	
  payoff	
  r(T)	
  =	
  wealth	
  (€)	
  to	
  be	
  
returned	
  to	
  the	
  customer
Portfolio (*) vs. Payoff
payoff r(T)
option price
How	
  to	
  adjust	
  dynamically	
  
the	
  portfolio	
  so	
  that	
  
wealthy(T) = payoff	
  r(T)	
  ?	
  ...
portfolio wealth
wealth y(T)
0.05
0.15
0.25
time	
  (years)
0.35
0.45
Stock price at expiration
asset	
  price	
  at	
  expiration
37 /50
Control systems technology
A typical control system
1001
D/A converters
actuators
sensors
A/D converters
control unit
38 /50
Most used sensors and actuators in control systems
Sensors
temperature
pressure
level
velocity, position
acceleration
force (strain) / deformation
Actuators
electrical motors (DC, brushless, step)
valves
heaters
39 /50
Sensors and actuators in control systems
angular position
thermocouple
pressure sensor
brushless motor
strain gage
liquid flow sensor
valve
40 /50
Open-loop vs. closed-loop control
Closed-loop control (feedback control)
Measurements of the output variables are fed back to the process through the controller
Open-loop control (feedforward control)
The manipulated input variable is generated without measuring the output variable
41 /50
Control systems design
How to design a (modern) control system
Understand the automation problem:
Which variables can be manipulated by actuators ?
What are the output variables of interest ?
What should we measure ?
Which are the disturbances ?
Get a reliable simulation model
Get a simplified mathematical model of the main process dynamics
Use design techniques to synthesize the control algorithm
Test in simulation, validate on real process
Control design requires a dynamical model of the open-loop process
42 /50
Course contents
What you will learn in this course
The tools to study the mathematical properties of dynamical
systems (control theory is often called “systems theory”)
Building simple dynamical models of common physical
processes
Computer-aided tools for analysis, simulation, and control of
dynamical systems (in MATLAB™)
43 /50
Analysis of dynamical systems in continuous time
(differential equations, Laplace transform, stability)
Models of dynamical systems (electrical, mechanical, hydraulic,...)
Analysis of dynamical systems in discrete-time
(difference equations, Zeta transform)
State-feedback control synthesis in the time domain (pole-placement)
State estimation
Output feedback control (dynamic compensator)
Integral action
Automatic Control 2: Frequency-domain analysis and synthesis (loop shaping)
and more advanced control techniques (optimal, nonlinear, predictive, ...)
44 /50
Textbooks
Suggested references
A. Bemporad - Lecture notes (what you’re looking at right now …)
(maybe enough …)
K.J. Åström, B. Wittenmark, Computer-controlled Systems,
Theory and Design, Prentice-Hall
(good classical textbook on digital control)
G.F. Franklin, J.D. Powell, M. Workman “Digital Control of
Dynamic Systems”, Addison-Wesley Longman.
T. Kailath, “Linear Systems”, Prentice-Hall
(advanced textbook for state-space concepts)
45 /50
E. Fornasini, G. Marchesini, “Appunti di Teoria dei Sistemi” ,
Edizioni Libreria Progetto (in Italian)
P Bolzern, R. Scattolini, N. Schiavoni, “Fondamenti di
controlli automatici” (in Italian)
(good textbook for frequency-domain)
K.J. Åström, R.M. Murray, “Feedback Systems: An Introduction
for Scientists and Engineers”, (available on-line for download)
(new textbook, plenty of examples)
46 /50
Logistics
Midterm exam: April 21, 2011. No expiration date.
= Final exam of Automatic Control 1 (TLC)
= Exam of AC2 for those elder mechatronic students that already
passed AC1.
Final exam: sometimes in June 2011 (AC1 + AC2).
Note: Submitting a new AC1 test automatically kills the midterm test !
Classroom exercises
Classroom exercises in MATLAB™
47 /50
Web site
http://www.ing.unitn.it/~bemporad/teaching/ac
30 novembre 2010
Lectures
Wednesday
10.30
13.00 Room A102
Thursday
Friday
Appointments for questions & explanations: send an email
Teaching assistants:
Matteo Rubagotti
Sergio Trimboli
http://www.ing.unitn.it/~rubagott
http://www.ing.unitn.it/~trimboli
49 /50
End of lecture
Questions ???
50 /50