WEBVTT
Kind: captions
Language: en
00:00:00.860 --> 00:00:06.890
Let us start getting round up with dynamics
and nonlinear systems. Now let start with
00:00:06.890 --> 00:00:15.849
the word dynamics, here is a simple definition,
it is the study of systems that evolve in
00:00:15.849 --> 00:00:25.330
time. Now such systems may eventually settle
down to some equilibrium or they may keep
00:00:25.330 --> 00:00:34.020
repeating in cycles or actually do something
much more complicated. Let us now visualise
00:00:34.020 --> 00:00:41.010
some dynamics, x is the depended variable,
time is the independent variable.
00:00:41.010 --> 00:00:48.530
There is an equilibrium value and their trajectories
settled to that equilibrium. You can have
00:00:48.530 --> 00:00:53.720
another scenario, where they do not settled
equilibrium; they actually just keep repeating
00:00:53.720 --> 00:00:59.899
themselves in cycles. The third scenario is
where you neither have an equilibrium nor
00:00:59.899 --> 00:01:07.470
a cycle but you keep doing something which
is fairly complicated.
00:01:07.470 --> 00:01:12.830
Now let us take a historical perspective on
dynamics the story, really started in mid
00:01:12.830 --> 00:01:24.200
1600s with Newton. What he did was he invented
calculus and he worked on differential equations.
00:01:24.200 --> 00:01:33.369
He then discovered his laws of motion and
universal gravitation and he combined them
00:01:33.369 --> 00:01:40.159
to explain Kepler’s laws of planetary motion.
00:01:40.159 --> 00:01:47.600
Now essentially what he did was solved two
body problem, which is calculating the motion
00:01:47.600 --> 00:02:00.939
of the earth around the Sun given the inverse
square law of gravitational attraction between
00:02:00.939 --> 00:02:07.869
them. Now what proved to be rather difficult
was the three body problem. That is the problem
00:02:07.869 --> 00:02:16.250
of the Sun, the Earth and the Moon, this turned
out to be very difficult problem to solve,
00:02:16.250 --> 00:02:26.220
in the sense of trying to obtain explicit
formulas for the motions of Sun, the Earth
00:02:26.220 --> 00:02:29.540
and the Moon.
00:02:29.540 --> 00:02:36.709
Progress then happened in late 1800s with
the work of Poincare. What he really did was
00:02:36.709 --> 00:02:44.370
introduce a brand new point of view, what
he said was the following: Let focus on qualitative
00:02:44.370 --> 00:02:52.480
rather than quantitative questions. Here is
an example of a quantitative question, can
00:02:52.480 --> 00:03:00.760
we find the exact positions of the planets,
at all times.
00:03:00.760 --> 00:03:15.129
Here is a qualitative question, will our solar
system be stable for ever, in other words
00:03:15.129 --> 00:03:24.430
might some of the planets suddenly decide
to leave the solar system and go off to another
00:03:24.430 --> 00:03:33.800
galaxy. Now what Poincare did was developed
a geometric approach to answer such qualitative
00:03:33.800 --> 00:03:34.960
questions.
00:03:34.960 --> 00:03:45.430
Now let us move on to the first half of 20th
century. Dynamics was largely concerned with
00:03:45.430 --> 00:03:54.640
nonlinear oscillators with applications in
physics and in engineering. Now this led to
00:03:54.640 --> 00:04:04.690
the development of the Radio, Radar, Phase
Locked Loop and Lasers. In the second half
00:04:04.690 --> 00:04:15.890
of the 20th century computers started to allow
us to visualise complex dynamics. Kepler Lorenz
00:04:15.890 --> 00:04:24.469
in his famous 1963 paper observed chaotic
motion in a simple model for atmospheric dynamics.
00:04:24.469 --> 00:04:33.520
Later in the 1970s, working in mathematical
biology, Winfree worked on nonlinear oscillators
00:04:33.520 --> 00:04:45.430
like heart rhythms. In physics Feigen Baum
went on to establish connection between chaos
00:04:45.430 --> 00:04:54.060
and phase transitions. And in the 1980s there
was a famous book by Kuramoto called chemical
00:04:54.060 --> 00:05:04.369
oscillations, waves and turbulence. Now moving
on to the beginning of the 21st century, the
00:05:04.369 --> 00:05:12.159
key words that one often hear are complex
systems and network systems.
00:05:12.159 --> 00:05:19.199
Now usually such systems are large scale and
highly nonlinear and their applications all
00:05:19.199 --> 00:05:24.189
over science and engineering.
00:05:24.189 --> 00:05:31.960
Now we start talking about dynamical systems.
There are two main types of dynamical systems,
00:05:31.960 --> 00:05:42.710
one is differential equations where the evolution
of the system happens in continuous time.
00:05:42.710 --> 00:05:51.809
The other variant is difference equations,
where time is discreet. We focus on differential
00:05:51.809 --> 00:06:01.599
equations, as these are widely used in science
and engineering. Very general frame work for
00:06:01.599 --> 00:06:15.059
ordinary differential equations abbreviated
as ODEs is x1 dot is equal to f1 x1 to x n
00:06:15.059 --> 00:06:27.849
all the way up to x n dot is equal to fn x1
to x n the over dot’s represent differentiation
00:06:27.849 --> 00:06:36.099
with the respect to time t so the xi dot is
equal to dxi dt.
00:06:36.099 --> 00:06:49.050
The variables x1 to x n could be concentrations
of chemicals within a reactor or there could
00:06:49.050 --> 00:07:00.260
be populations of different species within
an ecosystem and so on. The functions f1 to
00:07:00.260 --> 00:07:07.979
fn are determined by the specific problem
that had.
00:07:07.979 --> 00:07:14.639
As an example, let us consider Damped Harmonic
oscillator here is the differential equations
00:07:14.639 --> 00:07:27.209
that is m d square x dt squared plus b dx
dt plus kx is equal to zero, where m b and
00:07:27.209 --> 00:07:36.860
k are greater than zero. Now this is an example
of an ordinary differential equation. What
00:07:36.860 --> 00:07:48.449
we do is we introduce new variables x1 is
equal to x and x2 is equal to x dot. Then
00:07:48.449 --> 00:07:56.770
x1 dot is equal to x2 and x2 dot is equal
to x double dot which is equal to minus b
00:07:56.770 --> 00:08:11.550
on m x dot minus k on m x which is minus b
on m x2 minus k on m x1.
00:08:11.550 --> 00:08:23.580
The equivalent system when turns out be x1
dot is equal to x2 and x2 dot is equal to
00:08:23.580 --> 00:08:38.560
minus b on m x2 minus on k on m x1. Now recall
the general form of the ordinary differential
00:08:38.560 --> 00:08:45.060
equation that is x1 dot is equal to f1is the
function of x1 and x2 and x2 dot is equal
00:08:45.060 --> 00:08:55.560
to f2 is the function of x1 and x2 . So if
f1 is x2 and f2 is minus b on m x2 minus k
00:08:55.560 --> 00:09:04.459
on m x1 then we essentially just turn the
Damped Harmonic oscillator into general form
00:09:04.459 --> 00:09:06.829
of the ODEs that we had earlier.
00:09:06.829 --> 00:09:16.610
And here are some notes the above system is
linear, as all the xi’s on the right hand
00:09:16.610 --> 00:09:30.089
side are to the first power only. For a nonlinear
system, typically the terms are products,
00:09:30.089 --> 00:09:45.560
powers and functions of the xi’s. For example,
x1 x2, x1 squared sine of x1 etc are all examples
00:09:45.560 --> 00:09:50.190
of nonlinear terms.
00:09:50.190 --> 00:09:57.060
Now let us consider the example of a swinging
pendulum the differential equation is x double
00:09:57.060 --> 00:10:10.040
dot plus g on L sine x is equal to zero. Where
x is the angle of the pendulum from vertical,
00:10:10.040 --> 00:10:21.009
g is the acceleration due to gravity and L
is the length of the pendulum. The equivalent
00:10:21.009 --> 00:10:32.550
system is x1 dot is equal to x2 and x2 dot
is equal to minus g on L sine of x1. Now we
00:10:32.550 --> 00:10:39.980
get to this equivalent system using exactly
the same trick as the previous example.
00:10:39.980 --> 00:10:51.920
This system is nonlinear because of the sine
x1 term. This makes the equation very difficult
00:10:51.920 --> 00:11:01.389
to solve analytically. Now we can use the
small angle approximation that is sine x is
00:11:01.389 --> 00:11:09.899
approximately equal to x for x much less than
one, which will turn the equation into linear
00:11:09.899 --> 00:11:20.430
one. But we will also try to extract information
about the original nonlinear system using
00:11:20.430 --> 00:11:25.029
geometric methods.
00:11:25.029 --> 00:11:36.260
The rough idea is as follows now suppose that
for a particular initial condition, we just
00:11:36.260 --> 00:11:50.020
happen to know the solution to the pendulum
system. The solution could be a pair of functions
00:11:50.020 --> 00:12:05.360
x1 and x2 representing the position and velocity
of the pendulum. Now if we construct an abstract
00:12:05.360 --> 00:12:19.689
space x1, x2 then the solution x1 t and x2
t will correspond to a point moving along
00:12:19.689 --> 00:12:22.040
the curve in this space.
00:12:22.040 --> 00:12:29.069
Now let just go ahead and plot this we have
an initial condition x1 zero and x2 zero we
00:12:29.069 --> 00:12:45.870
go ahead and plot curve. Now this curve is
called a trajectory. Space is called the Phase
00:12:45.870 --> 00:12:55.940
Space. Now the Phase Space is actually filled
with trajectories, as each point can actually
00:12:55.940 --> 00:13:00.890
serve as an initial condition.
00:13:00.890 --> 00:13:11.300
Now we will outline the Poincare’s geometric
perspective. Now given the system we want
00:13:11.300 --> 00:13:22.959
to draw the trajectories and there by extract
information about the solutions. Now in numerous
00:13:22.959 --> 00:13:35.639
cases such geometric reasoning actually allows
us to draw trajectories without actually solving
00:13:35.639 --> 00:13:45.519
the system. Here are some notes the phase
space for the general system is the space
00:13:45.519 --> 00:13:51.459
with coordinates x1 to xn.
00:13:51.459 --> 00:14:03.180
So n represents the dimension of the phase
space, so the space is n dimensional. Now
00:14:03.180 --> 00:14:15.180
phase portrait is collection of all the qualitatively
different trajectories in the system.
00:14:15.180 --> 00:14:23.860
Now let us have brief discussion on linear
versus nonlinear systems. When you look at
00:14:23.860 --> 00:14:37.009
linear systems they can be broken into parts
then each part can be solved separately and
00:14:37.009 --> 00:14:49.579
then recombined to get to final answer. Now
they are equal to the sum of their parts and
00:14:49.579 --> 00:15:00.730
utilise methods like Laplace transforms and
Fourier analysis. So natural question to ask
00:15:00.730 --> 00:15:14.240
is given that we have such rich theory for
linear systems and very well established methods
00:15:14.240 --> 00:15:22.930
and they behave nicely then why not just model
using linear systems.
00:15:22.930 --> 00:15:35.430
The answer as follows when parts of the system
interfere, cooperate and compete. We very
00:15:35.430 --> 00:15:49.399
naturally get nonlinear interactions. Additionally,
most systems in science and engineering do
00:15:49.399 --> 00:15:53.879
not really act like linear systems.
00:15:53.879 --> 00:16:02.899
Now here is one way to look at world of dynamics.
On one axis we write down the dimension of
00:16:02.899 --> 00:16:11.779
the phase space and on the other axis we classify
system as linear or nonlinear. Then our current
00:16:11.779 --> 00:16:19.019
focus will be on this regime where we will
be dealing with one two dimensional nonlinear
00:16:19.019 --> 00:16:19.759
systems.