Lecture 2 - Our First Control Design 1.

Review

K. J. Åström • Feedback
• Open and closed loop systems
1. Introduction
• Properties of feedback
2. Cruise Control
+ Reduce effects of process disturbances
3. Agenda for the Future
+ Makes system insensitive to process variations
4. Differential Equations + Stabilize an unstable system
5. Summary + Create well defined relations between output and
reference
Theme: Make a simple control design. Systems described by
- Risk for instability
ordinary differential equations.
• PID control

2. Cruise Control Process Model

A simplified mathematical model
dv
m + vd = F − mgθ
dt

With reasonable parameters

dv
+ 0.02v = u − 10θ
dt

• Process input or control variable: gaspedal u where

• Process output: velocity v v [m/s] (10 m/s=36 km/h=22 miles/hour)
• Desired output or reference signal v r u normalized throttle 0 ≤ u ≤ 1
• Disturbances: slope θ θ slope in [rad/s]

Block diagram

c K. J. Åström September, 2000 1
 The Closed Loop System Stop and Think!!
dv How is the behavior of the equation
Process model: + 0.02v = u − 10θ
dt

t d2 e de dθ
PI controller: u = k(v r − v) + ki 0 (vr − v(τ ))dτ + (0.02 + k) + ki e = 10
dt2 dt dt
The closed loop system is described by (differentiate both influenced by the parameters k and k i
equations and add them) e = v r − v

d2 e de dθ
+ (0.02 + k) + ki e = 10
dt2 dt dt The Audience is Thinking ...
In steady state with constant θ and e we have e = 0.
No surprise the controller has integral action!

Behavior of the Equation How to Find Controller Parameters?

d2 e de dθ d2 e de dθ
+ (0.02 + k) + ki e = 10 2
+ (0.02 + k) + ki e = 10
dt2 dt dt dt dt dt

Compare with spring-mass-damper system Comparison with the normalized mass-spring-damper system

d2 x dx d2 x dx
m +d + kx = 0 + 2ζ ω 0 + ω 20 x = 0
dt2 dt dt2 dt

What are the effects of damping d and spring constant k? gives

Normalized parameters
k = 2ζ ω 0 − 0.02
2
d x dx ki = ω 20
2
+ 2ζ ω 0 + ω 20 x = 0
dt dt
 Reasonable to choose ζ = 1 critical damping. How to choose
d ω 0?
where ω 0 = k
m
and ζ = 0.5 √ .
mk

c K. J. Åström September, 2000 2
 ω = 0.1, ζ = 0.5 (dotted), ζ = 1 (solid), and ζ = 2 (dashed)
Comparison of Open Loop and Closed Loop? 2.5

Controller parameters ζ = 1, ω 0 = 0.1, open loop (dashed)

2

e [m/s]
1.5

closed (solid) when the road has a slope of 4% (10 m/s=36 1

km/h=22 miles/hour) 0.5

0
18
−0.5
0 10 20 30 40 50 60 70 80 90 100
16 t [s]
Velocity error [m/s]

14 0.5

12 0.4

u [0-1]
0.3
10

0.2
8
0.1
6
0
0 10 20 30 40 50 60 70 80 90 100
4 t [s]
2

0
0 10 20 30 40 50 60 70 80 90 100
Time [s]

ζ = 1, ω 0 = 0.05 (dotted), ω 0 = 0.1 (solid) and ω 0 = 0.2 ζ = 1, ω 0 = 0.05 (dotted), ω 0 = 0.1 (solid) and ω 0 = 0.2
(dashed) (dashed) with extra dynamics T = 5s
3 4

2.5
3
e [m/s]

e [m/s]
2
2
1.5
1
1
0
0.5

0 −1
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
t [s] t [s]
0.5 1

0.4 0.8
u [0-1]

u [0-1]
0.3 0.6

0.2 0.4

0.1 0.2

0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
t [s] t [s]


c K. J. Åström September, 2000 3
 What have we done? Agenda
• Derive process model in the form of ODE We must develop methods for
• ODE = Ordinary Differential Equation • Deriving equations for the system
• Select controller ODE • Methods for manipulating the equations
• Process and controller described in similar ways ODE • Techniques to understand the equations
• Eliminate variables to give relation disturbance output – Qualitative understanding concepts
• Understand how closed loop system behaves – Insight
• Select controller parameters to give desired behavior – Standard forms
• Fine tune parameters by simulation or experiment – Computations
• Methods for finding controller parameters
• Techniques to validate the results

A First Order Equation Parameter a tells a lot!

The homogeneous equation  t
y( t) = y(0) e−at
+b e− a( t−τ )u(τ ) dt
dy 0
+ ay = 0
dt
First term depends on initial conditions, second term depends
has the solution on input signal
y( t) = Ce− at a>0 10
a<0
1
8
where C = y(0) is an arbitrary constant. The equation 0.8
6
0.6
y

y
dy
+ a y = bu
4
0.4
dt 2
0.2

has the solution 0

0 1 2 3 4
0
0 0.5 1 1.5 2
 t
at at
y( t) = y(0) e−at + b e− a( t−τ )u(τ ) dt
0

c K. J. Åström September, 2000 4
 Equations of Higher Order Roots of Characteristic Equation give Insight!
Consider the equation A real root s = α to the characteristic equation corresponds to
the time function eα t .
dn y d n− 1 y d n− 2 y
+ a1 n−1 + a2 n−2 . . . + an y = 0
dt n dt dt Complex roots s = σ ± iω corresponds to the time functions.
The characteristic polynomial is eσ t sin ω t, eσ t cos ω t

A( s) = sn + a1 sn−1 + a2 sn−2 + . . . + an σ = 0.25ω σ = −0.25ω

40
0.6
If A(α ) = 0 then y(t) = eα t is a solution!. If the characteristic 0.4
30

equation has distinct roots α k the solution is

20

0.2

y
10


n 0 0
y( t) = Ck eα k t −0.2 −10
k=1
−0.4 −20
0 5 10 15 0 5 10 15

Roots of the characteristic equation gives insight! ωt ωt

Multiple Roots Ordinary Differential Equations

dn y d n− 1 y d n− 2 y dn y d n− 1 y
+ a1 + a2 + . . . + an y = 0 + a1 + . . . + an y = u
dtn dtn−1 dtn−2 dtn dtn−1

If α is a root to the characteristic polynomial has the solution


n 
A( s) = sn + a1 sn−1 + a2 sn−2 + . . . + an
t
y( t) = Ck−1 ( t) eα kt + h( t − τ )u(τ ) dτ
k=1 0
of multiplicity k and C(t) a polynomial of degree k − 1 then
y = C(t) eα t is a solution to the equation. Where h is the solution to the homogeneous equation
The general solution is dn h d n− 1 h
n
+ a1 n−1 + . . . + an h = 0
dt dt

n
y( t) = y0 ( t) = Ck−1( t) eα kt
with initial conditions
k=1

h(0) = 0, h (0) = 0, . . . , h n−2(0) = 0, h n−1(0) = 1


c K. J. Åström September, 2000 5
 General Linear Time Invariant System (LTI) Standard Forms
• Standard forms are foundations of the “language”
dn y d n− 1 y d n− 1 u
+ a1 + . . . + a n y = b1 + . . . + bn u • Learn to deal with the standard forms
dtn dtn−1 dtn−1
• Transform problems to standard form
has the solution
 • Software is often based on standard form

n t
α kt
y( t) = Ck−1 ( t) e + g( t − τ )u(τ ) dτ
k=1 0 One of the standard forms of LTI systems is

where Ck (t) are polynomials and the impulse response g is dn y d n− 1 y d n− 1 u

+ a1 + . . . + a n y = b1 + . . . + bn u
dtn dtn−1 dtn−1
g( t) = b1 h n−1( t) + b2 h n−2( t) + . . . + bn h( t)
A high order linear time invariant differential equation that
where h is the solution to the homogeneous differential equa- relates the output y of the system to its input u.
tion with initial conditions

h(0) = 0, h (0) = 0, . . . , h n−2(0) = 0, h n−1(0) = 1

Poles and Zeros Interpretation of Zeros

The linear time invariant system Consider the linear time invariant system
dn y d n− 1 y d n− 1 u dn y d n− 1 y d n− 1 u
+ a1 + . . . + a n y = b1 + . . . + bn u + a1 + . . . + a n y = b1 + . . . + bn u
dtn dtn−1 dtn−1 dtn dtn−1 dtn−1

is characterized by two polynomials Introduce

A( s) = sn + a1 sn−1 + a2 sn−2 + . . . + an−1 s + an A( s) = sn + a1 sn−1 + a2 sn−2 + . . . + an−1 s + an
B ( s) = b1 sn−1 + b2 sn−2 + . . . + bn−1 s + bn B ( s) = b1 sn−1 + b2 sn−2 + . . . + bn−1 s + bn

The roots of A(s) are called poles of the system. The roots of If s = β is a zero and u(t) = Ce β t it follows that
B (s) are called zeros of the system.
d n− 1 u d n− 2 u
The poles give the components of the time functions that b1 + b2 . . . + bn u = B ( β ) Ceβ t = 0
dtn−1 dtn−2
compose the solution.
A zero of B (s) at s = β blocks the transmission of the signal
u(t) = Ceβ t .

c K. J. Åström September, 2000 6
 The Transfer Function Inverse System
Consider the linear time invariant system Consider the linear time invariant system
dn y d n− 1 y d n− 1 u dn y d n− 1 y d n− 1 u d n− 2 u
n
+ a1 n−1 + . . . + an y = b1 n−1 + . . . + bn u n
+ a1 n−1 + . . . + an y = b1 n−1 + b2 n−2 + . . . + bn u
dt dt dt dt dt dt dt

Introduce the polynomials Introduce

A( s) = sn + a1 sn−1 + . . . + an−1 s + an A( s) = sn + a1 sn−1 + a2 sn−2 + . . . + an−1 s + an
B ( s) = b1 sn−1 + b2 sn−2 + . . . + bn−1 s + bn B ( s) = b1 sn−1 + b2 sn−2 + . . . + bn−1 s + bn

The rational function Notice (almost) symmetry between y and u. The inverse
system is obtained by reversing the roles of input and output.
B ( s)
B ( s)
A( s ) The transfer function of the system is and the inverse
A ( s)
is called the transfer function of the system A ( s)
system has the transfer function .
B ( s)

Cruise Control Simulation of LTI System

The relation between velocity error e = v r − v and the slope of Using the Control Systems Toolbox of Matlab it is very easy to
the road θ is described by the differential equation simulate an LTI system. Example cruise control

d2 e de dθ z=1.0;
+ (0.02 + k) + ki e = 10
dt2 dt dt w0=0.1;
t=0:0.01:100;
where k and k i are the controller parameters. This differential th0=0.04; %Slope of the road
equation is characterized by the polynomials k=2*z*w0-0.02;
A( s) = s2 + (0.02 + k) s + k i ki=w0^2;
B ( s) = 10s syseol=tf(10*th0,[1 0.02]);
sysecl=tf(10*th0*[1 0],[1 0.02+k ki]);
The poles can be given arbitrary values by choosing the eol=step(syseol,t);
controller parameters (pole placement). The choice k = ecl=step(syscl,t);
2ω 0 − 0.02 and k i = ω 02 gives a double pole at s = −ω 0. The plot(t,eol,’b--’,t,ecl,’b-’)
zero at s = 0 blocks transmission of constant θ .

c K. J. Åström September, 2000 7
 Summary
• Solution to a simple control problem:
– Write differential equations for process and controller
– Eliminate variables to obtain a differential equation for
closed loop
– Select controller coefficients so that the characteristic
equation has desired roots (or in another language the
closed loop system has given poles, pole placement).
• Review ordinary differential equations and the standard
model
dn y d n− 1 y d n− 1 u
+ a1 + . . . + a n y = b1 + . . . + bn u
dtn dtn−1 dtn−1

• Characteristic polynomial, poles, zeros and transfer

functions, inverse systems


c K. J. Åström September, 2000 8