Control Systems I

Lecture 2: Modeling and Linearization

Suggested Readings: Åström & Murray Ch. 2-3

Jacopo Tani

Institute for Dynamic Systems and Control

D-MAVT
ETH Zürich

September 28, 2018

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 1 / 37

Tentative schedule

# Date Topic
1 Sept. 21 Introduction, Signals and Systems
2 Sept. 28 Modeling and Linearization
3 Oct. 5 Analysis 1: Time response, Stability
4 Oct. 12 Analysis 2: Diagonalization, Modal coordinates
5 Oct. 19 Transfer functions 1: Definition and properties
6 Oct. 26 Transfer functions 2: Poles and Zeros
7 Nov. 2 Analysis of feedback systems 1: internal stabil-
ity, root locus
8 Nov. 9 Frequency response
9 Nov. 16 Analysis of feedback systems 2: the Nyquist
condition
10 Nov. 23 Specifications for feedback systems
11 Nov. 30 Loop Shaping
12 Dec. 7 PID control
13 Dec. 14 State feedback and Luenberger observers
14 Dec. 21 On Robustness and Implementation challenges

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 2 / 37

Today’s learning objectives

After today’s lecture, you should be able to:

Derive series, parallel and feedback system interconnections.

Understand the difference between the main control architectures (feedforward,

feedback).

Can use the concept of “state” to model a system.

Write the model of an LTI system with A, B, C, D matrices.

Understands the notion of equilibrium points and can calculate them.

The student is able to linearize a nonlinear system at an appropriately chosen

equilibrium point to derive an approximate LTI state space representation.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 3 / 37

Outline

1 Recap from previous lecture

2 Feedforward and feedback

3 State-space models

4 Modeling

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 4 / 37

Signals and Systems

A signal u is a map from a set, e.g. T, to another another set, e.g. W. In

short: u : T → W.
A system is a mapping between input and output signals.
Control systems uses block diagrams to represent systems and their
interconnections.

u y
Σ

We classify the system in different ways:

Static (memoryless) vs. Dynamic
Causal vs. Non-causal
Linear vs. Nonlinear
Time-invariant vs. Time-varying

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 5 / 37

Memoryless (or static) vs. dynamic systems

An input-output system Σ is memoryless (or static) if for all t ∈ T, y (t) is a

function of u(t).
In other words, in a static system the output at the present time depends
only on the value of input at the present time; not on the value of input in
the past or the future time.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 6 / 37

Causal vs. non-causal systems
An input-output system Σ is causal if, for any t ∈ T, the output at time t
depends only on the values of the input on (−∞, t]. It is strictly causal if the
output at time t depends only on values of the input (−∞, t).

In other words: a system is causal if and only if the future input does not affect
the present output.
All practically realizable systems are causal. (It is impossible to implement
a non causal system in “the real world”.)
J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 7 / 37
Linear vs. nonlinear systems
An input-output system is linear if it is additive and homogeneous.
Additivity: Σ(u1 + u2 ) = Σu1 + Σu2
Homogeneity: Σ(ku) = kΣu.
In other words, Σ is linear if, for all input signals ua , ub , and scalars α, β ∈ R,
Σ(αua + βub ) = α(Σua ) + β(Σub ) = αya + βyb .

ua ya ub yb
Σ Σ

αua αya
Σ

The key idea is superposition property:

αua + βub αya + βyb

Σ

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 8 / 37

Time-invariant vs. time-variant systems
An input-output system Σ is time-invariant if Σστ u = στ Σu = στ y , ∀τ ∈ T,
where στ is a shift operator such that στ u(t) = u(t − τ ).
In other words, the system manipulates the input in a way that does not
depend on when the system is used.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 9 / 37

Which systems will we treat in this course, and why?

In this course, we will consider only LTI SISO systems:

Linear,
Time Invariant,
Single Input, Single Output,
Causal.

This is a very restrictive class of systems; in fact, most systems are NOT
LTI. On the other hand, many systems are approximated very well by LTI
models. This is a key idea that we will explore today.

As long as we are mindful of the errors induced by the LTI approximation, the
methods discussed in the class are very powerful.

Indeed, most control systems in operation are designed according to the

principles that will be covered in the course.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 10 / 37

Interconnections of systems: series

Serial interconnection: u y
Σ1 Σ2
Σ = Σ2 Σ1

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 11 / 37

Interconnections of systems: parallel
u y
Σ1
Parallel interconnection:

Σ = Σ1 + Σ2 Σ2

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 12 / 37

Interconnections of systems: (negative) feedback
u y
(Negative) Feedback Σ1
interconnection: −

Σ = (I + Σ1 Σ2 )−1 Σ1 Σ2

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 13 / 37

Outline

1 Recap from previous lecture

2 Feedforward and feedback

3 State-space models

4 Modeling

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 14 / 37

Control objectives

Stabilization: make sure the system does not “blow up” (i.e., it stays within
“normal” operating conditions).

Regulation: maintain a desired operating point in spite of disturbances.

Tracking: follow a reference trajectory that changes over time, as closely as

possible.

Robustness: the control system satisfies the above objectives even if the
system is “slightly” different than we expected.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 15 / 37

The ideal controller, or, how to make a brick wall fly?

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 16 / 37

The ideal controller, or, how to make a brick wall fly?

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 16 / 37

The ideal controller, or, how to make a brick wall fly?

Example from Prof. Sandipan Mishra’s, “Feedforward and Iterative Learning Control”, RPI, 2012.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 16 / 37

Basic control architectures: Feed-Forward

r u y
F P

Basic idea: given r , attempt to compute what should be the control input u
that would make y = r .

Essentially, F should be an “inverse” of P.

Relies on good knowledge of P — sensitive to modeling errors.

Cannot alter the dynamics of P, i.e., cannot make an unstable system stable.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 17 / 37

Basic control architectures: Feedback

r e u y
C P
−

Basic idea: given the error r − y , compute u such that the error is “small.”

Intuition: the bigger C , the smaller the error e will be, regardless of P
(under some assumption, e.g., closed-loop stability).

Does not require a precise knowledge of P — robust to modeling errors.

Can stabilize unstable systems. But can also make stable systems unstable (!).

Needs an “error” to develop in order to figure out the appropriate control.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 18 / 37

Basic control architectures: two degrees of freedom

r e u y
C P
−

Basic idea: given r , compute a guess for the control input u that would make
y = r ; correct this guess based on the measurement of the error e, so that the
error is minimized.

Combines the main advantages of feed-forward and feedback architectures.

Ensures stability, robustness, but can also speed up tracking/regulation.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 19 / 37

Outline

1 Recap from previous lecture

2 Feedforward and feedback

3 State-space models

4 Modeling

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 20 / 37

Causality revisited
An input-output system Σ is causal if, for any t1 ∈ T, the output at time t1
depends only on the values of the input on (−∞, t1 ].

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 21 / 37

State of a system

We know that, if a system is causal, in order to compute its output at a given

time t0 , we need to know “only” the input signal over (−∞, t0 ].

This is a lot of information. Can we summarize it with something more man-

ageable?

Definition (state)
The state x(t0 ) of a causal system at time t0 is the information needed, together
with the input u between times t0 and t1 , to uniquely predict the output at time
t1 , for all t1 ≥ t0 .

Usually, the state of a system is a vector in some Euclidean space Rn .

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 22 / 37

State of a system

The state of the system at a given time t0 summarizes the whole history of
the past inputs over (−∞, t0 ), for the purpose of predicting the output at
future times.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 22 / 37

State of a system

The state of the system at a given time t0 summarizes the whole history of
the past inputs over (−∞, t0 ), for the purpose of predicting the output at
future times.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 22 / 37

State of a system

The state of the system at a given time t0 summarizes the whole history of
the past inputs over (−∞, t0 ), for the purpose of predicting the output at
future times.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 22 / 37

Water tank example

Suppose that the input is u = u+ − u− (i.e., net flow of water into the tank).

Is there a way to “summarize” the effect of all past inputs until time t?

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 23 / 37

Dimension of a system

The choice of a state for a system is not unique (in fact, there are infinite
choices, or realizations).

However, there are some choices of state which are preferable to others; in
particular, we can look at “minimal” realizations.

Definition (Dimension of a system)

The dimension of a causal system is the minimal number of variables sufficient to
describe the system’s state (i.e., the dimension of the smallest state vector).

We will deal mostly with finite-dimensional systems, i.e., systems which can
be described with a finite number of variables.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 24 / 37

LTI State-space model
It can be shown that any finite-dimensional LTI system can be described by a
state-space model of the form:

Definition (Continuous-time State-Space Models)

d
x(t) = Ax(t) + Bu(t);
dt
y (t) = Cx(t) + Du(t);

The state x represents the memory of the system.

The state x is an “internal” variable that cannot be accessed directly, but only
“controlled” though the input u and “observed” through the output y .
The order of the system is the dimension of the state x.
A system is memoryless if the “dimension” of the state is zero (i.e., there is
no need to keep a state, or memory)
A system is strictly causal if the “feedthrough” term D is zero.
J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 25 / 37
Outline

1 Recap from previous lecture

2 Feedforward and feedback

3 State-space models

4 Modeling

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 26 / 37

Water Tank (Integrator)

Note that: “All models are wrong”, but some are useful. (depending on the
application) – George E. P. Box.

Notation:
Flow: u = u+ − u− ,
Water level: h,
Tank cross-section area: Tcs ,
Dynamic model:
dh 1
= u
dt Tcs

State-space model, with y = x = h:

0 T1cs
   
A B
=
C D 1 0

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 27 / 37

Temperature Control (1st order system)

Notation:
Heat Flow: u,
Temperature: T ,
Outside Temperature: Te ,
Thermal Conductivity: c
Thermal Capacity: m
Dynamic model:
dT
m = c(Te − T ) + u
dt

State-space model, with

y = x = T − Te :

− mc 1
   
A B m
=
C D 1 0

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 28 / 37

Nonlinear systems

Unfortunately, most real-world systems are nonlinear.

Finite-dimensional, time-invariant, causal nonlinear control systems, with input
u and output y can typically be modeled using a set of differential equations
as follows:

Definition (Continuous-time Nonlinear State-Space Models)

d
ẋ(t) := x(t) = f (x(t), u(t)) ,
dt
y (t) = g (x(t), u(t));

Can we construct an LTI approximation of this system, so that we can apply

the techniques designed for LTI systems?
Remarkably, control systems designed for the LTI approximation will work very
well for the nonlinear system.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 29 / 37

Pendulum (2nd-order nonlinear system)

https://www.youtube.com/watch?v=oWiuSp6qAPk
Notation:
Angular position: θ,
Torque input: u,
Pendulum length: L,
Pendulum mass: m,
Viscous damping coefficient: c
Dynamic model:

mL2 θ̈ = −c θ̇ − mgL sin θ + u.

Define x1 := θ, and x2 := θ̇.

Then one can write

ẋ1 = x2 ,
c g
ẋ2 = − x2 − sin(x1 ) + u.
mL2 L

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 30 / 37

Equilibrium point

The pendulum is a nonlinear system. How can we construct an approximate

model which is linear?

The key idea is that of constructing an approximation that is valid “near” some
special operating condition, e.g., equilibrium points.

Definition (Equilibrium point)

A system described by an ODE ẋ(t) = f (x(t), u(t)) has an equilibrium point
(xe , ue ) if f (xe , ue ) = 0.

Clearly, if x(0) = xe , and u(t) = ue for all t ≥ 0, then

x(t) = x(0) = xe , for all t ≥ 0.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 31 / 37

Equilibria for a pendulum

What would be an equilibrium for the pendulum? We need to solve for

ẋ1 = x2 = 0,
c g
ẋ2 = − 2 x2 − sin(x1 ) + u = 0.
mL L

There are two solutions:

xe = (0, 0), ue = 0, and

xe = (π, 0), ue = 0.
The first equilibrium is stable, the second unstable; we will make these notions
more precise.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 32 / 37

(Jacobian) Linearization procedure

Given an equilibrium point (xe , ue ), with output ye = g (xe , ue ), a linearized

model is obtained by setting

x ← xe + δx,
u ← ue + δu,
y ← ye + δy ,

and then neglecting all terms of second (or higher) order in δx, δu, and δy in
the (nonlinear) dynamics model.

Note that since xe is a constant, ẋ = δ ẋ.

The resulting model would be a good approximation of the original nonlinear

model when “near” the equilibrium point, i.e., for |δx| << 1, |δu| << 1, and
|δy | << 1.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 33 / 37

Linearization about the stable equilibrium

Susbstituting as in the previous slide, with xe = (0, 0) and ue = 0,

δ ẋ1 = 0 + δx2 ,
c g 1
δ ẋ2 = − 2 (0 + δx2 ) − (0 + δx1 ) + (0 + δu),
mL L mL2
δy = δx1 .

The above can be rewritten in the state-space form as

δ ẋ = Aδx + Bδu,
δy = C δx + Dδu,

where    
0 1 0
A= , B= .
−g /L −c/(mL2 ) 1/(mL2 )
 
C= 1 0 , D = 0.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 34 / 37

Linearization about the unstable equilibrium

Susbstituting xe = (π, 0) and ue = 0,

δ ẋ1 = π + δx2 ,
c g 1
δ ẋ2 = − 2 (0 + δx2 ) − (0 − δx1 ) + (0 + δu),
mL L mL2
δy = δx1 .

The above can be rewritten in the state-space form as

δ ẋ = Aδx + Bδu,
δy = C δx + Dδu,

where    
0 1 0
A= , B= .
+g /L −c/(mL2 ) 1/(mL2 )
 
C= 1 0 , D = 0.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 35 / 37

Some remarks on infinite-dimensional systems

Even though we will not address infinite-dimensional systems in detail, some

examples are very useful:
Time-delay systems: Consider the very simple time delay ST , defined as a
continuous-time system such that its input and outputs are related by

y (t) = u(t − T ).

In order to predict the output at times after t, the knowledge of the input for
all times in (t − T , t] is necessary.
PDE-driven systems: Many systems in engineering, arising, e.g., in structural
control and flow control applications, can only be described exactly using a con-
tinuum of state variables (stress, displacement, pressure, temperature, etc.).
These are infinite-dimensional systems.
In order to deal with infinite-dimensional systems, approximate discrete models are
often used to reduce the dimension of the state.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 36 / 37

Today’s learning objectives

After today’s lecture, you should be able to:

Derive series, parallel and feedback system interconnections.

Understand the difference between the main control architectures (feedforward,

feedback).

Can use the concept of “state” to model a system.

Write the model of an LTI system with A, B, C, D matrices.

Understands the notion of equilibrium points and can calculate them.

The student is able to linearize a nonlinear system at an appropriately chosen

equilibrium point to derive an approximate LTI state space representation.

J. Tani, E. Frazzoli (ETH) Lecture 2: Control Systems I 09/28/2018 37 / 37