Nonlinear systems, chaos

and control in Engineering

Module 1
One-dimensional systems

Cristina Masoller
[email protected]
http://www.fisica.edu.uy/~cris/
 Schedule

Flows on the line Bifurcations Flows on the

(Strogatz ch.1 & 2) (Strogatz ch. 3) circle
18/11 (3 hs) 18/11 (2 hs) & (Strogatz ch. 4)
 Introduction 24/11 (3 hs) 25/11 (2 hs)
 Solving  Introduction  Introduction to
equations with  Saddle-node phase
computer  Transcritical oscillators
 Fixed points and  Pitchfork  Nonlinear
linear stability oscillator
 Examples
 Feedback control  Fireflies and
and delays entrainment
 Flows on the line: outline

 Introduction to dynamical systems

 Introduction to flows on the line
 Solving equations with computer
 Fixed points and linear stability
 Feedback control: delay differential equations
 Dynamical Systems

 Systems that evolve in time.

 Examples:
• Pendulum clock
• Neuron
 Dynamical systems can be:
• linear or nonlinear (harmonic
oscillator – pendulum); Voltage
• deterministic or stochastic;
• low or high dimensional;
• continuous time or discrete
time. Time
In this course: nonlinear systems (Nonlinear Dynamics)
 A neuron: an stochastic nonlinear
dynamical system

Koch Nature1997
 Given the initial condition:
possible evolution

• The neuron settles down to

equilibrium (rest state or “fixed
point”). – module 1 in this course

• Keeps spiking in cycles (“limit

cycle”).
– module 2

• More complicated: chaotic or

complex evolution (“chaotic
attractor”).
– module 3
Historical development of the
Theory of Dynamical Systems
 mid-1600s: Ordinary differential
equations (ODEs)

 Isaac Newton: studied planetary orbits and

solved analytically the “two-body” problem (earth
around the sun).

 Since then: a lot of effort for solving the “three-

body” problem (earth-sun-moon) – Impossible.
 Christiaan Huygens (mid-1600s,
Dutch mathematician)

 Patented the first pendulum clock.

 Observed the synchronization of two clocks.

http://www.youtube.com/watch?v=izy4a5er
om8
 Late 1800s

 Henri Poincare (French mathematician).

Instead of asking “which are the exact positions of planets
(trajectories)?”
he asked: “is the solar system stable for ever, or will planets
eventually run away?”

 He developed a geometrical approach to solve the problem.

 Introduced the concept of “phase space”.

 He also had an intuition of the possibility of chaos:

Poincare: “The evolution of a deterministic
system can be aperiodic, unpredictable, and
strongly depends on the initial conditions”

Deterministic system: the present state (initial

condition) fully determines the future state.
There is no randomness but the system can be
unpredictable.
 1950s: First simulations

 Computers drive economic growth and transform

how we live and work.
 Computes allowed to experiment with equations.
 Powerful tool to advance the “Theory of Dynamical
Systems”.
 1960s: Eduard Lorentz (American mathematician
and meteorologist at MIT): simple model of
convection rolls in the atmosphere.
 Intuition of chaotic motion on a strange attractor.
 He also showed that there is structure and order in
chaotic motion.
 The Lorentz system

Lorentz studied meteorological prediction using Navier-

Stokes simplified equations:
x   ( x  y )
y  x( r  z )  y
z  xy  bz
 3 variables:
• x: rotation rate of a cylindrical mass of gas,
• y: thermal gradient,
• z: temperature variation.
 3 Parameters:
• : ratio between viscosity and thermal conductivity (Prandtl number),
• R: temperature difference between top and bottom of cylinder (Rayleigh
number),
• b: ratio between width and height of the cylinder.
 Lorentz’s Attractor

 Starting from an initial condition (x0, y0, z0) by

numerically integrating the equations we can plot the
trajectory in the phase space (Lorentz’s Attractor).

Lorentz found
extreme
sensitivity to
initial conditions
 impossibility
of long-term
meteorological
predictions.
 Order within chaos and
self-organization

 Ilya Prigogine (Belgium, born in Moscow, Nobel

Prize in Chemistry 1977)
 Thermodynamic systems far from equilibrium.
 Discovered that, in chemical systems, the interplay
of (external) input of energy and dissipation can
lead to “self-organised” patterns.
 Reverse the rule of maximization of entropy
(second law of thermodynamics).
 Wide implications to biological systems and the evolution of
life.
 One-dimensional
spatio-temporal patterns

Vanag et al Nature 2000

Patterns in nature
(source: wikipedia)
 Entropy (H) and Complexity (C)

H=0 H≠0 H = Max

C=0 C≠0 C=0

O. A. Rosso (2009)
 According to Prigogine

Newtonian physics has been extended three times:

• First, with the use of the wave function in quantum

mechanics.

• Then, with the introduction of space-time in relativity.

• And finally, with the recognition of indeterminism in

nonlinear systems.

Chaos is the third great revolution of 20th-century

physics, after relativity and quantum theory.
 In the 1960s: biological
nonlinear oscillators

 Arthur Winfee (American theoretical biologist –

born in St. Petersburg): Large communities of
biological oscillators show a tendency to self-
organize in time –collective synchronization.

In the 1960’s he did experiments trying to understand the

effects of perturbations in biological clocks (circadian rhythms).

What is the effect of an external perturbation on

subsequent oscillations?
 Winfee work on perturbing
biological oscillators

 He studied the periodic emergence of a fruit fly that as a 24-

hour rhythmic emergency.

 Using brief pulses of light, found that the periodic emergence of

the flies was shifted, and the shift depended on the timing and
the duration of the light pulse.

 Also found that there is a critical timing and duration that results
in no further periodic emergency (destroys the biological clock).

 The work has wide implications, for example, for cardiac tissue:
some cardiac failures are related to perturbed oscillations.
 The 1970s

 Robert May (Australian, 1936): population biology

 "Simple mathematical models with very
complicated dynamics“, Nature (1976).
xt 1  f ( xt ) Example: f ( x)  r x(1  x)

 Difference equations (“iterated maps”), even though

simple and deterministic, can exhibit different types of
dynamical behaviors, from stable points, to a
bifurcating hierarchy of stable cycles, to apparently
random fluctuations.
 Models of dynamical systems

Continuous-time
ordinary differential Discrete-time equations
equations (ODEs) (iterated maps)
dx
 x  f ( x, r ) xt 1  f ( xt , r )
dt

r = control parameter(s)

Example: Lorentz model Example: Logistic map

x   ( x  y ) xt 1  r xt (1  xt )
y  x( r  z )  y
z  xy  bz

r = (, r, b)
 The logistic map
1
x(i  1)  r x(i )[1  x(i )]
r=2.8
r=2.8, Initial condition: x(1) = 0.2
x(i)

0.5
Transient relaxation → long-term stability
0
0 10 20 30 40 50
1 i
r=3.3
“period-doubling”
bifurcations to chaos
x(i)

0.5

0
0 10 20 30 40 50 Transient
i
1 r=3.5 dynamics x(i)
→ stationary
x(i)

0.5

0
oscillations
0 10 20 30 40 50
1
i (regular or
irregular)
x(i)

0.5
Parameter r
r=3.9
0
0 10 20 30 40 50
i
 Universal route to chaos

 In 1975, Mitchell Feigenbaum

(American mathematical physicist), using
a small HP-65 calculator, discovered the
scaling law of the bifurcation points
rn 1  rn  2
lim n  4.6692...
rn  rn 1
 Then, he provided a mathematical proof (by using the
“renormalization concept” –connecting to phase transitions in
statistical physics).
 Then, he showed that the same behavior, with the same
mathematical constant, occurs within a wide class of functions,
prior to the onset of chaos (universality).
Very different systems (in chemistry, biology, physics,
etc.) go to chaos in the same way, quantitatively.
 HP-65 calculator

 The first magnetic card-programmable handheld

calculator

HP-65 in original hard

case with manuals,
software "Standard
Pac" of magnetic
cards, soft leather
case, and charger
 The late 1970s

 Benoit Mandelbrot (Polish-born, French

and American mathematician 1924-
2010): “self-similarity” and fractal
objects:
each part of the object is like the whole
object but smaller.

 Because of his access to IBM's

computers, Mandelbrot was one of the
first to use computer graphics to create
and display fractal geometric images.
 Fractal objects

 Are characterized by a “fractal” dimension that measures

roughness (more in Module 3)

Romanesco broccoli Human lung Coastline of Ireland

D=2.66 D=2.97 D=1.22

Video: http://www.ted.com/talks/benoit_mandelbrot_fractals_the_art_of_roughness#t-149180
 In the 80’s: can we observe
chaos experimentally?

 Optical chaos: first observed in laser systems.

Time
…

N. B. Abraham, OPN 1989 More latter about experiments

with optical chaos in our lab.
 In the 90’: can we control
chaotic dynamics?

 Ott, Grebogi and Yorke (1990)

Unstable periodic orbits can be used for control: wisely
chosen periodic kicks can maintain the system near the
desired orbit.

 Pyragas (1992)
Control by using a continuous self-controlling feedback
signal, whose intensity is practically zero when the system
evolves close to the desired periodic orbit but increases
when it drifts away.
 Experimental demonstration of
control of optical chaos

 Raj Roy and others (1994)

 The 1990s: synchronization of two
chaotic systems
Pecora and Carroll, PRL 1990

Unidirectionaly coupled Lorenz systems: the ‘x’

variable of the response system is replaced by the
‘x’ variable of the drive system.

 Drive system

 Response system
 Interesting but … useful?

 Transmission of secure information

(Cuomo–Oppenheim, PRL 1993)

Is the system secure?

 In the case of low dimensional chaotic attractors, is
possible to break the system and extract the message
(G. Perez y H. Cerdeira, PRL 1995)
 Different type of coupling and different
types of synchronization

• Complete (CS): x1(t) = x2(t) (identical systems)

• Phase(PS): the phases of the oscillations are

synchronized, but the amplitudes are not.

• Lag (LS): x1(t+to) = x2(t)

• Generalized (GS): x2(t) = f(x1(t)) (f depends on

the strength of the coupling)
Experimental observation with
coupled lasers
Fischer et al Phys. Rev. A 2000
 Synchronization with time-delay

 Solution: 𝑥 𝑡 = 𝑦(𝑡 −)  𝑥(𝑡 +) = 𝑦(𝑡)

Is it possible to anticipate/predict
the evolution of a chaotic system?

The solution is stable only for ‘small’ values of  (Voss, PRL 2001)
Experimental demonstration
with electronic circuits
M. Cizak et al, PRL 2003
 In the last decade: can we
exploit / use chaotic dynamics?

An example from optical chaos

After processing the signal, ultra-fast generation of

arbitrarily long sequences of random bits.

I. Kanter et al, Nature Photonics 2010

39
 Dynamical systems in optics:
Two examples from our lab
 Laser spikes: intensity vs. time  Neuronal spikes

https://youtu.be/nltBQG_IIWQ

 Extreme pulses
Intensity time series

Interested? TFGs available, contact us.

From dynamical systems to
complex systems
 End of 90’s - present

 Interest moves from chaotic systems to complex systems

(low vs. very large number of variables).

 Networks (or graphs) of interconnected systems

 Complexity science: dynamics of emergent properties

• Epidemics
• Rumor spreading
• Transport networks
• Financial crises
• Brain diseases
• Etc.
 Synchronization of a large
number of coupled oscillators

http://
www.y
outub
e.com
/watch
?v=D
D7YD
yF6dU
k
 Kuramoto model
(Japanese physicist, 1975)
 Model of all-to-all coupled phase oscillators.
d i K N

dt
 i 
N
 sin(
j 1
j  i )  i , i  1...N

K = coupling strength, i = stochastic term (noise)

Describes the emergence of collective behavior (synchronization)

How to quantify synchronization? N
1
With the order parameter: rei   e j
i

N j 1
r =0 incoherent state (oscillators are scattered in the unit circle)
r =1 all oscillators are in phase (i=j  i,j)
 Synchronization transition as the
coupling strength increases

video

Strogatz
Nature 2001
 Synchronization transition

 The Kuramoto model allows for deriving exact

results (Steven Strogatz and others, late 90’).

Video: https://www.ted.com/talks/steven_strogatz_on_sync
 Network science

The challenge: to understand how the interplay of

structure and dynamics (of individual units) results
in emergent collective behavior.

Strogatz
Nature 2001,
 Main feature of a network: the
degree distribution

Liu et al, Nature 2011

Scientific coorperation: red
FISES (A. Diaz-Guilera, UB)
Electric power grid (N. Rubido
PhD Thesis 2014)
New York State electric power
grid

Generators and substations

are shown as small blue bars.
The lines connecting them
are transmission lines and
transformers. Line thickness
and colour indicate the
voltage level: red, 765 kV and
500 kV; brown, 345 kV;
green, 230 kV; grey, 138 kV
and below. Pink dashed lines
are transformers.
Strogatz, Nature 2001
Graph consisting of the pages of a web site and their
mutual hyperlinks, which are directed. Communities
are indicated by the colors (Fortunato, 2010)
Nodes represent
symbiotically
connected species,
such as plants and
pollinators (Gao et
al, Nature 2016)

Interacting
networks
(example:
friendship
relations in
Facebook and
twitter)
Transport networks
 The start of Network Theory: The
Königsberg Bridge problem
The problem: to
devise a walk
through the city that
would cross each
bridge once and only
once. The starting
and ending points of
the walk need not be
the same.

By representing the network as a Graph (a set of “vertices”

connected by a set of “edges”) Euler (1707-1783) proved
that the problem has no solution.
 Brain functional network

Eguiluz et al, PRL 2005

27/11/2016
Chavez et al, PRE 2008
Nodes and links of the
climate network

57
 Time series of a climate variable (air Similarity measure (correlation,
temperature, wind, precipitation, etc.) mutual information, etc.)

Winds & ocean currents Threshold

Climate network

Eguiluz et al, PRL 2005

Deza et al, Chaos 2013
Donges et al, Chaos 2015
 Summary and take home
message

 The theory of dynamical systems allows to

• understand the dynamics of low-dimensional systems,
• to uncover “order within chaos”,
• universality features in the transition to chaos, and
• provides tools for controlling chaotic behavior.

 Complexity science is aimed to understand the emerging

phenomena in large sets of interacting systems.

 Both, dynamical systems and complexity science have

multiple applications, and involve the work of mathematicians,
physicists, biologists, computer scientists, engineers, etc.
 Outline

 Introduction to dynamical systems

 Introduction to flows on the line
 Solving equations with computer
 Fixed points and linear stability
 Feedback control: delay differential equations
 Types of dynamical systems

 Continuous time: differential equations

• Ordinary differential equations (ODEs).

Example: damped oscillator

• Partial differential equations (PDEs).

Example: heat equation

 Discrete time: difference equations or “iterated

maps”. Example: the logistic map
x(i+1)=r x(i)[1-x(i)]
 ODEs can be written as first-
order differential equations

 x  f (x)

 First example: harmonic oscillator

 Second example: pendulum


 Trajectory in the phase space

 Given the initial conditions, x1(0) and x2(0),

we predict the evolution of the system by
solving the equations: x1(t) and x2(t).
 x1(t) and x2(t) are solutions of the equations.
 The evolution of
the system can be
represented as a
trajectory in the
phase space.
 two-dimensional
(2D) dynamical
system. Key argument (Poincare): find out
how the trajectories look like, without
solving the equations explicitly.
 Classification of dynamical systems
described by ODEs (I/II)
x  f ( x)   (t )

 f(x) linear: in the function f, x appears to first order only

(no x2, x1x2, sin(x) etc.). Then, the behavior can be
understood from the sum of its parts.
 f(x) nonlinear: superposition principle fails!
 Example of linear system: harmonic oscillator
In the right-hand-side x1

and x2 appear to first
power (no products etc.)

 Example of nonlinear system: pendulum


 Classification of dynamical systems
described by ODEs (II/II)
x  f ( x)   (t )
 =0: deterministic.
 0: stochastic (real life) –simplest case: additive noise.
 x: vector with few variables (n<4): low dimensional.
 x: vector with many variables: high dimensional.
 f does not depend on time: autonomous system.
 f depends on time: non-autonomous system.
 Example of non-autonomous
system: a forced oscillator

 Can also be written as first-order ODE

 3D system: to predict the future evolution we

need to know the present state (t, x, dx/dt).
 So…what is a “flow on the line”?

 A one-dimensional autonomous dynamical

system described by a first-order ordinary
differential equation

 x 
 f does not depend on time
 Summarizing

Number of variables
N>>1 N= (PDEs
N=1 N=2 N=3
DDEs)
• Heat
RC circuit Harmonic equation,
Linear oscillator • Maxwell
equations
• Schrodinger
equation

• Navier-
Logistic Pendulum • Forced • Kuramoto
Nonlinear Stokes
population oscillator phase
(turbulence)
grow • Lorentz oscillators
model

PDEs=partial differential eqs.

“flow on the line” DDEs=delay differential eqs.
 Outline

 Introduction to dynamical systems

 Introduction to flows on the line
 Solving equations with computer
 Fixed points and linear stability
 Feedback control: delay differential equations
 Numerical integration

 Euler method

 Euler second order

 Fourth order (Runge-Kutta 1905)

 Problem if t is too small: round-off errors

(computers have finite accuracy).
 Example 1

• quiver(x,y,u,v,scale): plots y  y(1  y )

arrows with components (u,v) 2
at the location (x,y). 1.8

1.6

• The length of the arrows is 1.4

scale times the norm of the 1.2

(u,v) vector.

y(t)
1

0.8

To plot the blue arrows: 0.6

0.4
%vector_field.m 0.2
n=15;
0
tpts = linspace(0,10,n); 0 2 4
t
6 8 10

ypts = linspace(0,2,n);
[t,y] = meshgrid(tpts,ypts);
pt = ones(size(y));
py = y.*(1-y);
quiver(t,y,pt,py,1);
xlim([0 10]), ylim([0 2])
 Numerical solution

y  y(1  y ) y(0)  0.1

2
To plot the solution (in red): 1.8

1.6

1.4

tspan = [0 10]; 1.2

yzero = 0.1;

y(t)
1

[t, y] =ode45(@myf,tspan,yzero); 0.8

plot(t,y,'r*--'); xlabel t; ylabel y(t) 0.6

0.4

0.2

0
function yprime = myf(t,y) 0 2 4 6 8 10
t
yprime = y.*(1-y);
The solution is always tangent to the arrows

Remember: HOLD to plot together the blue

arrows & the trajectory.
 Example 2

y   y  5et sin 5t y(0)  0.5

1.5

n=15;
1
tpts = linspace(0,3,n);
ypts = linspace(-1.5,1.5,n); 0.5
[t,y] = meshgrid(tpts,ypts);
pt = ones(size(y));

y(t)
0

py = -y-5*exp(-t).*sin(5*t);
-0.5
quiver(t,y,pt,py,1);
xlim ([0 3.2]), ylim([-1.5 1.5]) -1

-1.5
0 0.5 1 1.5 2 2.5 3
tspan = [0 3]; t

yzero = -0.5;
function yprime = myf(t,y)
[t, y] = ode45(@myf,tspan,yzero);
yprime = -y -5*exp(-t)*sin(5*t);
plot(t,y,'kv--'); xlabel t; ylabel y(t)
General form of a call to Ode45
Class and homework

0
10

-5
10

-10
10

-15
10
-4 -2 0
10 10 10
 Outline

 Introduction to dynamical systems

 Introduction to flows on the line
 Solving equations with computer
 Fixed points and linear stability
 Feedback control: delay differential equations
 Example

Analytical Solution:

 Starting from x0=/4, what is the long-term behavior (what

happens when t?)

 And for any arbitrary condition xo?

 We look at the “phase portrait”: geometrically, picture of all

possible trajectories (without solving the ODE analytically).

 Imagine: x is the position of an imaginary particle restricted to

move in the line, and dx/dt is its velocity.
 Imaginary particle moving in the
horizontal axis
Flow to the right when x  0
Flow to the left when x  0 x0 =/4

x0 arbitrary

x  0 “Fixed points”

Two types of FPs: stable & unstable

 Fixed points

Fixed points = equilibrium solutions

 Stable (attractor or sink): nearby

trajectories are attracted
 and -

 Unstable: nearby trajectories are

repelled
0 and  2
 Example 1

 Find the fixed points and classify their stability

Example 2
 Example 3: population model for
single species (e.g., bacteria)

 N(t): size of the population of the species at time t

 Simplest model (Thomas Malthus 1798): no migration,

births and deaths are proportional to the size of the
population

Exponential grow!
 More realistic model:
logistic equation

 To account for limited food (Verhulst 1838):

 If N>K the population decreases

 If N<K the population increases

 K = “carrying capacity”
 The carrying capacity of a biological species in an
environment is the maximum population size of the species
that the environment can sustain indefinitely, given the food,
habitat, water, etc.
How does a population approach
the carrying capacity?

 Exponential or sigmoid approach.

 Good model only for simple

organisms that live in constant
environments.
 And the human population?

Hyperbolic grow !
Technological advance
→ increase in the carrying
capacity of land for people
→ demographic growth
→ more people
→ more potential inventors
→ acceleration of
technological advance
→ accelerating growth of
the carrying capacity…

Source: wikipedia
 Linearization close to a
fixed point
 = tiny perturbation

Taylor expansion

The slope f’(x*) at the fixed point determines the stability

the perturbation  grows exponentially
the perturbation  decays exponentially
Second-order terms can not be neglected and a
nonlinear stability analysis is needed.
Bifurcation (more latter)
Characteristic time-scale
 Existence and uniqueness

 Problem: f ’(0) infinite

 When the solution of dx/dt = f(x) with x(0) = x0 exists and is

unique?
 Short answer: if f(x) is “well behaved”, then a solution exists
and is unique.
 “well behaved”?
 f(x) and f ’(x) are both continuous on an interval of x-values
and that x0 is a point in the interval.
 Details: see Strogartz section 2.3.
 Example 1

 Linear stability of the fixed points of

 Stable:  and -

 Unstable: 0,  2
 Example 2

 Logistic equation

The two fixed points have

the same characteristic
time-scale:
Good agreement with controlled
population experiments
 Lack of oscillations

General observation: only

sigmoidal or exponential
behavior, the approach is
monotonic, no oscillations

Analogy:

Strong damping
(over damped limit)

To observe oscillations we need

to keep the second derivative
(weak damping).
 Stability of the fixed point x*
when f ’(x*)=0?

In all these systems:

When f’(x*) = 0
nothing can be
concluded
from the
linearization
but these plots
allow to see
what goes on.
 Potentials

V(t) decreases along the trajectory.

 Example:

Two fixed points: x=1 and x=-1

(Bistability).
 Summary

 Flows on the line = first-order ODE: dx/dt = f(x)

 Fixed point solutions: f(x*) =0

• stable if f´(x*) <0
• unstable if f´(x*) >0
• neutral (bifurcation point) if f´(x*) = 0

 There are no periodic solutions; the approach to a fixed

point is monotonic (sigmoidal or exponential).
 Class and homework

10

7
x (0)
x (t )  6

1  2tx 2 (0)
5

-1
0 2 4 6 8 10
 Outline

 Introduction to dynamical systems

 Introduction to flows on the line
 Solving equations with computer
 Fixed points and linear stability
 Feedback control: delay differential
equations
 Exception to no oscillations:
delay differential equations (DDEs)

 Any system involving a feedback control will almost

certainly involve time delays.
 In a 2D system delayed feedback can reduce oscillations,
but in a 1D system it can induce oscillations.
 Example:

 Linear system
 Infinite-dimensional system
 Delay-induced oscillations.
 Example: population dynamics
Delayed logistic equation

The initial function is y=0.5 in -1<s<0

 In a single-species population, the incorporation of a delay allows to

explain the oscillations, without the predatory interaction of other species.
 Example: Container crane
Delayed feedback control
It is important for the crane to move payloads rapidly and
smoothly. If the gantry moves too fast the payload may start
to sway, and it is possible for the crane operator to lose
control of the payload.
 Pendulum model for the crane,
y represents the angle

weakly damped oscillator Feedback control

(not first-order equation,
without control payload
oscillations are possible)

Reduction of payload oscillations: why delayed feedback works?

Near the equilibrium solution y=0:

Small delay:

The delay increases the damping.

Therefore: the oscillations decay faster.
 but

 Small perturbation  Large perturbation

Example: Car following model
 Typical solution for two cars

The lead vehicle reduces its speed of 80 km/h to 60

km/h and then accelerates back to its original
speed. The initial spacing between vehicles is 10 m.

 Speed of the two cars  Distance between the two cars

 Alcohol effect

A sober driver needs about 1 s in order to start

breaking in view of an obstacle.
With 0.5 g/l alcohol in blood (2 glasses of wine), this
reaction time is estimated to be about 1.5 s.
 oscillations near the stable equilibrium increase.
 Solving DDEs

Example 1: Delayed logistic equation

function solve_delay1
tau = 1; 2.5

ic =constant
ic = [0.5];
initial 2

tspan = [0 100];
function
h = 1.8; 1.5

sol = dde23(@f,tau,ic,tspan); 1

plot(sol.x,sol.y(1,:),'r-')
0.5
function v=f(t,y,Z)
v = [h*y(1).*(1-Z(1))]; 0
0 20 40 60 80 100

end
end
 Solving DDEs

Example 2: Prey (x) and predator (y) model

function solve_delay2
45
tau=9; 40

ic = [35;10]; 35

tspan = [0 250]; 30

25
h = 10; 20

sol = dde23(@f,tau,ic,tspan); 15

plot(sol.x,sol.y(1,:),'r-',sol.x,sol.y(2,:),'b--') 10

function v=f(t,y,Z) 0
0 50 100 150 200 250

v = [y(1)*(2*(1-y(1)/50)-y(2)/(y(1)+40))-h
y(2)*(-3+6*Z(1)/(Z(1)+40))];
end
end
 Bibliography

 Steven H. Strogatz: Nonlinear dynamics

and chaos, with applications to physics,
biology, chemistry and engineering.
First or second ed., Chapters 1 and 2

 Thomas Erneux: Applied delay differential

equations (Springer 2009).

 D. J. Higham and N. J. Higham, Matlab

Guide Second Edition (SIAM 2005)