Week 4 video transcripts (videos attached separately)

=======================================================

Lesson: Systems' Solutions and Behaviours

0:05
So now, we're going to talk about solution and behavior of a system. So let's say
we are given a system S, which contains this tuples and let's look at this
sequence. And I'm going to use this naming for this sequence, right? So I put u at
time 0, v at time 0, x at time 0 and y at time 0 all in a 4-tuple and I actually
concatenate for all this value at time from 0 til t-1. So that gives me a sequence
and I'm using this naming for this sequence. I simply use the tuple u, v, x, y to
denote this sequence. So now this sequence given as this is a solution of the
system S on the interval zero to T. So one thing I should emphasize here about this
notation. So look at the notation part in the bottom of this slide. So every time I
am talking about intervals in real numbers, I'm actually use comma to denote
intervals in real. So for example, these means a closed interval both from the left
and right is closed, close from the below and close from top. So here, it means
close from below and open from top. So here, it means open from below and close
from top right and this interval means open from both below and above.
Play video starting at :1:54 and follow transcript1:54
So we can also use infinity to use the upper bound and minus infinity to use lower
bound of the interval in real. So now if I'm interested to denote intervals on
integer when a and b are integer numbers, I actually use semicolon to denote
intervals over integer. So for example, this interval means integer numbers all the
way from a to b, including a and b. This means integer numbers from a to b
including a, but not including b. This means integers from a to b not including a,
but including b. And this means integers from a to b, but not including a and not
including b.
Play video starting at :2:50 and follow transcript2:50
Okay, let's come back to our notion of solution. So again, we are given a system S.
This sequence which is namely a sequence of four tuples containing u, v, x, y from
time 0 til time t-1 is called a solution of the system S over the interval from 0
to t. Again, this means integers from zero to capital T including zero, but not
including capital T which t can be a bounded integer or it can also be infinity. So
it's a solution if first of all, x at time zero has to be inside the initial state
set. X(t+1), of course, has to be inside F(x(t), v(t)). So that means a state at
the next side should actually fulfill the state transition map and this should hold
for all t from 0 til t-1 including 0, but not including t-1. And (y(t), v(t)) has
to be inside H(x(t), u(t)). So that means output and internal input variable should
also satisfy the conditions of the output map and this should also hold for all t
from zero til capital T including zero, but not including capital T. So if we have
a sequence of this form satisfying these three conditions which simply means
satisfying the property of the system definition, right? So x should satisfy the
state transition function dynamic and y and v should satisfy the output function
condition. So in that case, this sequence is simply called a solution of the
system. So we also have what is called behavior of the system. So when we talk
about behavior of the system, we only care about external input and output of the
system which are available from outside. We actually do not consider internal
variables when we are talking about behavior of the system. So keep in mind, when
we talk about solution of the system, we consider external input, internal input,
state, and output variables. But when we talk about behavior of the system,
behavior of the system only contains external input and output of the system
because these are the two entities which are available from outside. We can observe
them, we can measure them.
Play video starting at :6:1 and follow transcript6:01
So again, we use the same naming terminology. So if you look at the sequence of
pairs, so u(0), comma y(0), u1, comma y1 all the way til u(T-1), y(T-1) is inside
the behavior of the system S. So that means behavior of the system is actually a
set of sequences, right? If and only if there exists internal input variable v and
a state variable x and capital T such that u, comma v, comma x, comma y is a
solution of the system on the interval from zero tit T including zero, but not
including T. And if T is not infinity, that implies that the pair x(T-1), v(T-1) is
blocking. So let's understand what's going on. So, when we talk about behavior of
the system, behavior of the system is the set of all sequences of pairs of u, comma
y which are infinite. We only allow finite sequences to belong to behavior of the
system if the corresponding state internal input pair at the last given time is
blocking. That's the only way we allow finite sequences to be part of the behavior
of the system. If You look at the corresponding solution for a given T. And if the
pair x(T-1), v(T-1) is blocking, then in this case we have a finite behavior.
Otherwise, the elements of the behaviors of the system, they are all infinite
sequences of pairs of (u, y). So okay, let's assume I have this tuple (u, v, x, y)
be a solution of a system S on this interval over integer including 0, not
including T. So if T is less than infinity, so that means this solution itself is a
finite sequence, right?
Play video starting at :8:57 and follow transcript8:57
So now let's assume the system S is simple. So in this case, remember, if the
system S is simple, external input and internal inputs are the same, state and
output are also the same. So in this case, this tuple reduces to this. Therefore,
we can actually summarize, rather than writing as a four tuple. Because u is
repeating, x is repeating, we can actually use the sequences of pairs (u, x) to
also denote solutions of simple system. But now if you have a simple system, you
can see the structure of the solution and behaviors then become the same. Meaning
that both of them are sequences of pairs of (u, x). But there is a differences,
right? Solution can be finite and infinite. So each finite solution (u, x) of S
that ends with a blocking pair is also an element of the behavior. Each infinite
solution of S is in the behavior as well. So as I explained in the previous slide,
elements of the behavior of the system are all infinite sequences.
Play video starting at :10:29 and follow transcript10:29
We only allow for finite sequences to belong to behavior of the system.
Play video starting at :10:38 and follow transcript10:38
Only if we look at the corresponding solution and the corresponding pair of x and v
at time T- 1 is blocking. Otherwise the elements of the behavior are infinite
sequences. So now if we come back in this slide. So when we have simple systems, we
know that internal input and output, internal input and external input are the
same. And the state and output are also the same, right? So in this case, the tuple
representing the solution reduces to just a pair, (u, x), which syntactically is
the same as pairs we are using for behavior. However, solutions can be finite or
infinite. However, behaviors can only allow to be infinite, and they are only
allowed to be finite only if we end up at a blocking pair, right? And that's what
the explanation here is saying. Saying each finite solution which ends with a
blocking pair is also an element of the behavior. And each infinite solution is
automatically the element of the behavior if the system is simple.
Play video starting at :12:6 and follow transcript12:06
So now let's actually look at this specific example, and let's come up with some
samples of solutions and behaviors. First of all, let's look at the state diagram.
So, based on our convention, how do we draw a state diagram? From previous
lectures, we know that this represent a basic system. So before we actually write
the samples of solutions and behaviors, let's come up with the tuples
representation of the system, right? So in this case, what is the, State set? So
the state set is, we have two states, 1 and 2.
Play video starting at :12:52 and follow transcript12:52
What is the initial state set? Remember, we show initial state with the source less
arrow, which in this case is state number 1.
Play video starting at :13:4 and follow transcript13:04
So since this is a basic system, internal input and external input set are the
same. So let's only write the external input set, which is, in this case, we have
two input 1, 2.
Play video starting at :13:19 and follow transcript13:19
What about output set?
Play video starting at :13:21 and follow transcript13:21
So from the picture, we can already see output a, b. And based on this extra
information written here for blocking case, so we know that there is also another
output d.
Play video starting at :13:41 and follow transcript13:41
Okay, so now let's pick a particular interval, for example, [0; 3[. And in this
case, 3 is open, so means our T, when we define the notion of solution or behavior,
T is 3 in this case. So the length of the sequences for solutions has to be 3.
Okay, let's come up with a,
Play video starting at :14:15 and follow transcript14:15
Solution for this system of length 3. Okay, so remember, the solution was four
tuples, right? Sequences of four tuples. But in this case, I mean, in this case,
it's a basic system, right? So internal input and external input are the same, but
let's just write them, right? But you can also write in a succinct way and just
repeat the external input. And you don't need to repeat internal input because it's
the same as external one.
Play video starting at :14:50 and follow transcript14:50
Okay, so if we start from state 1, which is our initial condition, and if I apply
input, let's say 2, so input is 2, that's external input. Internal input is the
same as external input, and so what is the initial state, state 1.
Play video starting at :15:11 and follow transcript15:11
And what is the corresponding output? So if I'm in state 1, apply input 2, the
output is b.
Play video starting at :15:20 and follow transcript15:20
So then we go to state number 2. So let's say if you are in state number 2, we
apply input 1, what is the corresponding output, is a.
Play video starting at :15:34 and follow transcript15:34
So since we applied input 1, we stay on state 2.
Play video starting at :15:44 and follow transcript15:44
Here, this is a sample of a solution of this system of length 3.
Play video starting at :15:55 and follow transcript15:55
Okay, now let's come up with Finite and infinite behavior. First of all, does this
system has finite behaviors? Okay, in order to have finite behavior, we need to
have blocking pairs. Yes, this system has a blocking pairs, right? So look, F of if
I'm in a state two and if I apply input two, you see there is no outgoing
transitions. So that means pair two, comma two is a blocking pair. So in order to
have finite behaviors, so two, two has to be the last element of the behavior,
okay? So let's have a finite behavior. So here is an example of this was solution
of length three. Now I would like to come up with finite behavior, okay? So
remember, behaviors were sequences of pairs of u, comma y, okay? So initially, we
start from state one. So if I apply input one, the output is a.
Play video starting at :17:19 and follow transcript17:19
So if I apply input one, I stay there. And then if I apply input two, the
corresponding output will be b and then I go to a state two. If I'm in a state two,
again, I can apply input one which in this case, the corresponding output will be a
and I can actually take that self-loop for some time. So one, comma a. You can
continue this one, comma a. But then because I want to be finite behavior,
eventually, I apply input two and the corresponding output will be d. So this is
blocking and that's it, I stop. So you see, now I have a finite behavior, right?
And it's a valid behavior because you see, I end up with the input two at state two
is actually it results in a blocking pair.
Play video starting at :18:40 and follow transcript18:40
What if I want want to to have an infinite behavior? I mean, infinite behavior is
very easy. For example, this is a simple example. Start from state one and just
sustain that self-loop on a state one. Okay, just apply input one forever. This is
a correct infinite behavior. Okay, what if I would like to have a solution of size
four?
Play video starting at :19:11 and follow transcript19:11
Okay, I start from state one. Let's say we apply input one, one, comma one.
Play video starting at :19:28 and follow transcript19:28
So then I will be from a state one and the corresponding output is a. So I state in
state one and then let's apply input two, then I go to, I mean, given that the
input was one, I still stay on state one. And in this case, the corresponding
output is b. So since the input was two, I go to estate two. So let's say in a
state two, I apply input one. And so since I applied input one, I stay in a state
two. And let's say the next input, I also apply input one. So this is a solution of
length four.

Lesson: System Composition - Parallel Composition

0:06
So after discussing about solutions and behavior of the system, now we have all the
ingredients to actually talk about system compositions. How can we interconnect
systems and build new systems? So the first composition structure is called
parallel composition. Let's see, parallel composition is defined. So let's assume
we have two systems, system S1 and S2. So the parallel composition of S1 and S2 is
defined by we use this notation, these two vertical line to show parallel
composition of S1 and S2 is equal to a new system. And let's see how the elements
of the tuple of this new system is being defined. So in this case, the state set of
the parallel composition is simply Cartesian product of the state set of the two
system. The initial state set of the composition is simply again Cartesian product
of the initial state sets of two system. The external input set of the parallel
composition is also Cartesian product of the external input set of two system. The
internal input set of the parallel composition is also Cartesian product of
internal input sets of two systems. The output set of the parallel composition is
also Cartesian product of the output set of two systems. So let's see how the state
transition function and output function is defined. So in this case, the state
transition function is a function of the state set of the parallel composition,
internal input set of the parallel composition, and it's also set valued map. Its
co domain is a state set of the parallel composition. Let's see how F12 is defined.
So F12 of a state in the composition, internal input of the composition is simply
equal to the Cartesian product of a state transition function of system 1 and a
state transition function of system 2. So let's see how the output function is
defined. So again, as you recall, output function is a function of product of a
state and external input and its co-domain is product of output set and internal
input set. So in this case, H12 of (X,U) is simply Cartesian product of output
function of system 1, output function of system 2. So in nutshell, how this
intuitively, when we have two systems, the parallel composition of them simply
means put these two systems together and look at them as one system. That's how
parallel composition of two system is defined.

Lesson: System Composition - Serial Composition

0:05
The next interconnection structure we're going to talk about is called serial
composition. Let's assume we are given two system S_1 and S_2. S_1 is serial
composable with S_2 if the output set of System 1 is a subset equal of external
input set of System 2. Let's look at this picture. We are given System 1 and System
2. As you can readily verify yourself, you can see in order to have a well defined
serial interconnection, the output generated by System 1 should be acceptable as an
input to System 2. That's the reason we need to assume that the output set of
System 1 is subset equal of external input set of System 2 in order to have a well
defined serial interconnection. The serial composition of System 1 and 2 is defined
by, we use this o notation to denote serial interconnection of System 2 and System
1. When we write S_2 o S_1, that means S_1 is serially interconnected to S_2, which
means the output of System 1 goes as an input to System 2. The serial
interconnection of System 1 and 2 itself is a system. Let's look at how that
elements in this topple is define. The state set of serial interconnection is
simply Cartesian product of a state set of System 1 and 2. The initial state set of
the serial interconnection is also Cartesian product of initial state sets of
System 1 and 2. Look, the external input set of the serial interconnection is
simply the external input set of System 1, which is also can be readily seen from
this figure. Look, the input to the serial interconnection is external input to
System 1. The internal input set is also Cartesian product of internal input sets
of System 1 and 2. The output set is the output set of System 2. Look at the
figure, you see the output coming out of the serial interconnection is simply
output of System 2. Let's see how the state transition and output function are
defined. The state transition function is simply Cartesian product of a state
transition function of System 1, a state transition function of System 2. What
remains is we need to define how the output function is defined. As you know,
output function, its domain is Cartesian product of the state set and external
input set. Its co domain is Cartesian product of output set and internal input set.
Now let's see how the output function is defined. H_12 of state x, u_1 is equal to
the set of all the topples. First element is y_2 the output, internal input, which
is the pair of v_1, v_2 such that there exist y_1 one inside the output set of
System 1 such that y_1, v_1 is inside H_1 of x_1, u_1 and y_2, v_2 is inside H_2 of
x_2, y_1. That's how the output function of the serial interconnection of S_1 and
S_2 is defined. Let's look at an example. Let's assume System S is a simple system.
You see here I use the extended topple notion. But as you know for simple system we
can also work with the coin size representation of simple system. Meaning that we
don't need to write this ID that because the output function is identity and we
don't need to repeat internal input set because it's equal to external input set.
Let's assume Q is a static system.
Play video starting at :4:57 and follow transcript4:57
Internal and external input set is equal to the state set of System S. It's output
set is equal to X', so it's static system. We know how the F_q is defined, so
starting from q under any internal input x, we're going to be on q itself. So it's
a static system. Let's see how the output function of this static system is
defined. It says the output of a state q, external input x is equal to the H(x)*x
itself, for some strict map H.
Play video starting at :5:44 and follow transcript5:44
Now we would like to look at the serial interconnection of S and Q, which is given
by Q o S. You can actually imagine this Q is playing the role of quantizer, which
is quantizing the output of system S. Consider system S, its output goes inside the
static system Q. Here you can imagine Q being a quantizer. It's a static system.
This map edge is actually tells us how, in the case of simple system, the output is
equal to a state. How the states are quantized to the state set X prime. Now let's
look at their serial interconnection. As you recall from the previous slide, the
state set of the serial interconnection is Cartesian product of a state set. The
initial set also Cartesian product. The external input set is equal to the external
input set of system S. The internal input set is also Cartesian product of internal
input set. The output set is the output set of the static system, which is X prime.
Play video starting at :7:11 and follow transcript7:11
We already know that the state transition function is simply Cartesian product of
the state transition function, which is exactly these. The state transition
function of system S times the state transition function of static system, which is
Q itself. Let's see how the output function is defined. H12 of, if you're in a
state X, Q, and if the external input U, it's equal to, it's going to be what is
the output of the serial interconnection? Is going to be the output of static
system Q, which is H of X times the internal input which is the pair of U, X. Since
Q is a static system, we can actually reduce and make this representation much more
coin size. We can actually get rid of this Q. We don't need to write this Q because
the static system,
Play video starting at :8:23 and follow transcript8:23
the topple actually reduces to this. Which is containing a state set X, which is
the set of system S. Initial state set of system S, and external input set is equal
to the external input set of system S. What is the output set? Look, now we are not
dealing with a simple system anymore. Rather we are dealing with a basic system.
Why is that? Because the output of system S, which was simple, goes inside a
quantizer. Now we are dealing with a new output set, which is the output set of the
system Q which was X prime. The state transition function is going to be a state
transition function of the simple system. The output function is also going to be
different, meaning that H prime, if you're in a state X and external input, is U, H
prime of X, U is equal to H of X. It's not going to be equal to X anymore, rather
is going to be equal to H of X. This H was some strict map which was defining the
output function of the static system Q. This is an application of serial
composition. If you have a system and a quantizer comes after the system, you can
actually model this simply using serial composition of the two systems. In fact, in
many physical system, if you recall, we talked about sample and hold linear control
system, which was systems evolving in continuous time. Then we were able to come up
with a sample and hold version and model that as a simple system, if you recall
that. Now here we have discretize the measurements of the system with the sampling
time tall. Those measurements come, we call them X. Now X goes inside this static
quantizer Q and the quantize state information, X hat, comes out of it. Remember
the role of zero or the hold which used to transfer discrete input to continuous
time input for the system. Let's assume we have the quantizer Q. It takes N
dimensional vector X and provides N dimensional integer valued vector XR. Let's see
how those XR hat is defined. Q of X is simply equal to XR, which are integer valued
such that X minus XR, infinity norm less than equal to one. This is one way of
designing quantizer. This is just an example for the sake of illustration of serial
composition of a sample and hold linear control system with a quantizer. Now this
quantizer can be simply identified with the static system of this form, in which X
hat the output set of the quantizer is equal to N dimensional integer valued
vectors. Because it's static, the state transition function is simply a self-loop
over the state q itself. The output function is defined, the H of state q comma.
External input x is equal to the q of x times the x itself. This is the output, and
this is the internal input.
Play video starting at :12:18 and follow transcript12:18
Remember we described the sample and hold system using a simple system in which X =
X_0, equal to n-dimensional Euclidian space. U was in general dimensional Euclidian
space. We are interested in serial composition of S and Q. Q composed serially with
S it's actually equal to a basic system, which X, X_0, U are the sets defined for
system S. F is also the same being defined for system X and X hat is the output set
of the serial interconnection of system q and s which is equal to the output set of
the quantizer. Look at here the picture. It's equal to the output set of the
quantizer, which is X hat. H is the output function. Let's see. It's basic in which
H(X, U)= Q(x). That's how we can actually write the serial interconnection of this
sample and holding a control system with a quantizer as a basic system. In the
previous two example, we had a static system coming after the system, which was the
role of quantizer to quantize state information of the system. One can also put the
quantizer before the system and still look at the serial interconnection of the
quantizer system. Again, assume the system S is a simple system and Q is a static
system, but here now Q comes before system. Look at here. The internal and external
input sets are U prime, and the output set of Q is actually equal to the external
input set of system S. It's a static system. We know how the F_q is defined is just
a simple self-loop. In this case, the output function is Q of state Q, external
input U prime is equal to the output which is equal to H_u prime times the internal
input, which in this case internal input and external input are the same. H, here
is some strict set value map H. In this case, Q comes before the system S. Why is
that? Look at the definition of the output set of system Q. Output set of system Q
is equal to external input set of system S. This satisfies the posedness of serial
interconnection. Remember, the output set of system 1 should be sub equal to
external input set of system 2, which is the case here. The serial composition of Q
and S now is given by this system, which in this case state set is Cartesian
product of state. Initially, state set also Cartesian product. The external input
set is the external input set of system Q. Internal input set is the Cartesian
product of internal input sets of system Q and S. Output set is the output set of
simple system, which is X. The state transition function is simply Cartesian
product state transition function of Q and system S. Let's see how the output
function is defined. The output function apply on a state pair. Remember, in the
product, the states are pair Q, X, and an external input, U prime is equal to the
output, which is the output of system S, which is X itself. The state measure is a
simple system. The output is equal to the state times internal input. H_u prime is
the internal input system Q and U prime is the internal input system Q. Now one can
actually write this pick top representing the serial composition of S Q. The more
coincide manner we can actually get rid of those. In this case, the serial
composition of S and Q is simply the total. State set is the state set of system S.
Initial state set is the initial state set of system S, external input set is U
prime is equal to external input set of Q, internal input is U, output set is X. A
state transition function is a state transition function of system S and H prime of
X comma U prime is equal to, so the output is X and internal input is H of U prime.

Lesson: System Composition - Feedback Composition

0:06
Now we're going to actually finish the system composition with our last composition
structure, which is called feedback composition. Let's see how feedback composition
is defined. Assume we are given two systems, system S_1 and S_2. We say S_1 is
feedback composable with S_2 if the output set of system 2 is subset equal in
external input set of system 1, and the output set of system 1 is also subset equal
external input set of system 2. In addition, system 2 we consider to be Moore.
Remember system is called Moore if the output is not explicitly a function of
external input. If you look at this figure, you understand why the output set of
system 2 has to be subset equal external input set of system 1 and vice versa look.
This is how the feedback composition look like. The output of system 2 goes as an
input to system 1, and the output of system 1 goes as an input to system 2. This is
called feedback composition. That's the reason in order to have a well-posed
interconnection, we need to have this condition. I will explain later why we need
system 2 to be moved in the context of feedback composition. Now the feedback
composition of S_1 and S_2 is defined by and we use times sign to indicate feedback
composition of S_1 and S_2. Feedback composition of S_1 and S_2 itself is a system,
but what is very interesting to see is feedback composition of S_1 and S_2 the
external input set is singleton. Actually, even though S_1 and S_2 both they have
external input set, when you look at their feedback composition, the external input
set becomes singleton. You can also see that from this figure. You see when you
look at the feedback composition of S_1 and S_2, there is no more external input
available. The only thing we see is actually the output of the system. There is no
arrow going inside the feedback composition of S_1 and S_2. That's the reason the
input external input set is now singleton and the feedback composition is in fact
an autonomous system based on the definition of autonomous systems which we had
early on. Let's see how the elements of these topples are defined in the context of
feedback composition. The state set of the feedback composition is actually
Cartesian product of a state set of system 1 and 2. The same thing for initial S
state set. Initial S state set of the feedback composition is also Cartesian
product of initial S state set of system 1 and 2.
Play video starting at :3:29 and follow transcript3:29
External input set is we already said, it's a single tone. What about internal
input set of the feedback composition is the Cartesian product of internal input
set of system 1 and 2. What about output set? Output set is also Cartesian product
output set of system 1 and 2.
Play video starting at :3:54 and follow transcript3:54
Let's see how the state transition function of the composed system is defined.
Play video starting at :4:5 and follow transcript4:05
A state transition function of the composed system of state X, which is the
Cartesian product. State X here is a pair, X_1, X_2. The same thing with internal
input is also a pair, V_1, V_2 is simply Cartesian product of a state transition
function of system 1 F_1(x_1, v_1)* F2(x_2, v_2). Let's see how the output function
is defined.
Play video starting at :4:38 and follow transcript4:38
What is the domain of output function is only the state of the compose system
because we don't have any more external input set remember. In general, the output
function is a function of the state and external input set. In the context of
feedback composition, the external input set is singleton. Then the output function
of the compost is only a function of the state of the composition, which is X_12.
It's co domain is the product of the output set and internal input set. Now let's
see how the output function is defined. H_12 acting on a state of the compost. Here
X is a pair of x_1, x_2 is equal to pair of y_1, y_2. These are the output of the
compose system, pair of v_1, v_2. That's the internal input of the compose system.
Such that, (y_1, v_1) is inside H_1(x_1). Remember, in the feedback composition,
the output of System 2 goes as an input to System 1 and vice versa. That's the
reason here in the place of external input of System 1, we put y_2. The same thing
for System 2 as well, (y_2, v_2) inside belongs to H_2(x_2), external input of
System 2, which is in fact the output of System 1. We put y_1 here. That's how the
output function of the feedback composition of S_1 and S_2 is defined.
Play video starting at :6:42 and follow transcript6:42
Now let's look into why we ask S_2 to be Moore. The reason we are asking in the
context of feedback composition S_2 to be Moore because this actually prevents
circular dependency. What do I mean by circular dependency? Remember, y_2 is the
output of System 2, so goes as an input to System 1. We know that the output
function of System 1 depends on its external input, which is the output of System
2.
Play video starting at :7:27 and follow transcript7:27
Also on the other hand, we know that the output of System 1 gets affected based on
the output function H_1. Then we also know that y_1 goes as an input to System 2.
This means y_1 also through the map, H_2 is going to affect y_2. Now you see y_2 is
actually affecting y_2 itself. We have this circular dependency going on. To
prevent the circular dependency, we assume that System 2 is Moore, meaning what?
System 2 being Moore means its output does not explicitly depend on external input.
That means y_2 doesn't get affected explicitly based on y_1. This way we can
actually prevent the circular dependency. This is actually also the case based on
the real application. Usually in real application, S_1 is the controller and S_2 is
the plant. In those case actually S_2 when it's plant, the output of the plant does
not immediately get affected by the external input. In this case S_1 feedback
composition with S_2 is also called the closed loop. For those of you who are
familiar with linear control theory, that's what the role of S_1. S_1 is the
controller, S_2 is the plant, and their feedback composition is also called the
closed loop of the system. This picture shows the feedback composition of System 1
and 2, and you can readily see why the feedback composition of System 1 and 2 is an
autonomous system, meaning that the external input set is singleton. Because all
the external inputs are being used for the sake of interconnection, and there is
nothing available after the interconnection. Now, one question of interest is, is
S_1 feedback composed with S_2 itself a system? Remember, in order to be a system,
meaning that all the sets defining the system has to be non-empty. This has to be
non-empty. This has to be non-empty, non-empty, non-empty, non-empty. Now if you
look at the definition of X_12, X_12,0, V_12, and Y_12, you see they are not empty,
because system 1 and system 2 they were both system. So that means the
corresponding sets are not empty. Therefore, the sets defining the feedback
compositions are also non-empty.
Play video starting at :10:30 and follow transcript10:30
The other condition for a total definition to be a system was the output map needs
to be a strict. What does it mean the output map needs to be a strict? Meaning
that, for any x in this cartesian product, H_12 of X, we need to show it's non-
empty. That's what we mean by H_12 being a strict. So for any given X in the
composed system. For any X in this cartesian product, we need to show H_12 of X is
non-empty. This is also straightforward to show. Let's pick any X in the composed
system X inside X_12. Let's pick any X. Now what we're going to do, let's pick the
pair Y_2, V_2 belongs to H_2 of X_2, U_2. For some U_2 inside the external input
set of System 2. Let's pick a pair inside this set.
Play video starting at :11:51 and follow transcript11:51
Now we choose the pair Y_1, V_1 inside H_1 of X_1, Y_2. Remember the output of
System 2 goes as an input to System 1 and we already have Y_2 defined here, so
let's use this Y_2 fed it as an input to System 1. Now we have this pair, right?
Because System 1 was system, that means H_1 of X_1, Y_2 is non-empty, right? Then
we have a pair. Let's call that pair Y_1, V_1. Now, since S_2 is more, remember
based on definition of more, there should exist an internal input, V_2 prime in
System 2 such that Y_2, V_2 prime belongs to H_2 of X_2, Y_1. Now look, what I did
was I changed the external input of System 2 from U_2 to Y_1, which got defined
here in this line. Since the system is moved, we know that if we change the
external input, I should be able to still see the same output for some other choice
of internal input. Remember? Great. Now it follows that so now I know the pair Y_1,
V_1 belongs to H_1 X_1, Y_2 and Y_2, V_2 prime belongs to H_2 of X_2, Y_1. That
means this map is for any given X. This map is non-empty. We are able to show that
there is a corresponding pair of Y, V in the composed system, which belongs to H_ 1
two of X. Hence, the map is a strict, which implies the feedback composition of S_1
and S_2. It's in fact a system itself. Remember when I introduced the feedback
composition, I said usually System 2 is the plant System 1 is the controller, so
here let's look at this example. In this special case, System 2 is a simple system
S, and System 1 is a controller, we call it C and let's assume it's a static
system. For those of you familiar with linear control theory, you have actually got
exposed to a lot of static state feedback controller. Those are static systems.
Same thing here. C is a static system. And also if you look at S, and first of all
S is a simple system, hence it's more. What else? C is static. Let's see if the
feedback composition is well posed. In order to be well-posed, the external input
set of C needs to be subset equal of the output set of S which is the case, they're
actually the same.
Play video starting at :15:23 and follow transcript15:23
The output set of C which is U has to be subset equal external input set of S which
they're equal so then we satisfy all the conditions of feedback composition. Also S
is simple, hence it's more.
Play video starting at :15:41 and follow transcript15:41
Now, let's see how this output function of the static system is defined in this
particular example. It says, H_c(q,x), it says a function of a state and external
input is equal to its output is H(x) times so x itself is also internal input for
some strict H. Again, for those of you familiar with the feedback systems, linear
control theory, you know that this H in fact plays what we call is called state
feedback controller. This H is actually what we deal with usually in closed loop
systems when our controller is state feedback controller, so that's actually this H
plays that role. It gets the measurements from the system which are X and it
provides admissible input to the system such that the closed loop for example
satisfy some property of interest. For example, stability. Now, the feedback
composition of C, the static system C and the simple system S is given. What I did,
I just wrote the elements of the topple based on this slide. Here, we have
Cartesian product of the state set, Cartesian product of the initial state set, the
external input set is singleton, Cartesian product of internal input set, Cartesian
product of the output set and we need to see how the state transition function,
output function is defined. State transition function is also a Cartesian product
of a state transition function as you can see here. What about the output function?
The output function, actually you don't need to write this zero because here we
just simplify the notation because the external input set was singleton, we don't
need to write the H_1,2 be a function of external input because it's only
singleton. H_1,2 of the state is equal to the pair of output and pair of internal
input such that x equal to x prime, u equal to u prime, and u belongs to H(x). You
see that's how the output of the system state is, u belongs to H(x). Of course,
again, we can simplify this notation. We don't need to write state q because C is
static. We can just get rid of this q, we can get rid of this q, we can keep this
singleton and this is the external output set U x X, and the state transition
function is we only need to worry about the state transition function of system S
because C was static, it was only a simple [inaudible] over q. H prime, so where H
prime is defined as this H prime of x,0. Again, you don't need to write zero. The
same thing here. H prime of x is equal to the pair of the output, which is u,x such
that U belongs to H(x) and this H was on the strict map in the definition of the
static system C. Usually in reality, this H is actually what we call a state
feedback controller that gets the measurement, whether the measurements are state
value. State value comes as a measurement, goes inside the controller Controller
using this map edge calculate all possible inputs that actually can be applied to
system such that the closed loop satisfy some property of interest.