Advertisement
English transcript
Hello everybody. Welcome to this new course on optimal control guidance and estimation.
Typically, it will be covering around aerospace engineering problems and applications thing
like that, but largely the techniques, I mean tricks and techniques will be very generic.
So, anybody from any other discipline can also use this concept at equal level. I am
Radhakant Padhi working as a associate professor in Department of Aerospace Engineering, Indian
Institute of Science, Bangalore.
So, let us see some of this introduction, motivation and things like that, before we
go to the details of the course. This is the main objective of this particular lecture
actually. So, before we start begin the some points to note. The topics covered in this
course are very generic with preferential emphasis on aerospace applications obviously,
because largely if we see this optimum control applications have been given by aerospace
engineering tragic optimum problems conventionally. And we also want to kind of develop the course
and this course is meant to equip the students with applied optimal control system and state
estimation concepts. So, as part of it, we will also take this missile guidance and other
equal missile guidance problems as well, which is nothing but an application of optimal control
theory, and from that prospective, we will discuss actually.
And the topics covered as I told are fairly mathematical; that means somebody who loves
mathematics can, can, enjoy it very well, but if you, if you are reasonably good in
mathematics, I think you will be able to follow it well, but that is a necessary requirement
of this particular course. And I also suggest all of you that, if you that as the course
goes along, you take a set of paper, pencil and all use start kind of practicing everything
yourself, then only the geometry comes here more actually.
So, the topics to be covered in this particular course is a first thing is a some review of
necessary fundamentals; that means we will talk about some numerical methods, matrix
theoretical, matrix algebra and whatever necessary background material, that, that, will be quickly
review. Then followed by static optimization concepts and that is a kind of a backbone
for optimal control in away, because even if we see one, one, method of solution of
optimal control is, is, first to convert it to a static optimization problem and then
take the advantage of static optimization and rotation all that. So, that we need to
know a little bit of a static optimization before we go ahead with the, with optimal
control principle, which is nothing but dynamic optimization.
So then, we will go to calculus of variation and see some of the concepts of. So, what
is this calculus of variation; obviously, it is a, it is a vast topic; we are not able
to cover everything, but whatever is relevant to us, I think that much we will we able to
overview before proceeding to calculus of variation application for optimal control
and that is what we will talk next optimal control through calculus of variation, and
then, that will be followed by some numerical techniques to solve this variation and calculus
problem, which essentially leads to these something call two point one revalue problem.
We will talk about that is the course develop actually.
But that is one of the reasons why it is computationally very kind of very challenging and computationally
over well and rather, and then, we will proceed for other techniques, reason techniques and,
and, classical as well as recent techniques to how to overcome some of those actually.
Anyway, after covering numerical techniques, we will move to this something called LQR
theory or linear quadratic regulator theory and that will be followed by some overview
of a flight dynamics, because most of the concepts you want to use in aerospace application.
So, it is better to know some of these flight dynamics on the way.
Then we will see how LQR is use for flight control applications. This is some of the
application like stability turbulent let us and then many variable announcement and think
like that. We will talk about that as an application of flow domain. Next their application LQR
will be optimal missile guidance and that is what lot of people do using linear system,
system, dynamics rather engagement dynamics between vehicle and target. That is what they
do. The missile and target, if you take relative motion and all the using that as a constraint
and then formulate some sort of a minimum problem, and then, solve it using l. So, those
details will talk after that. That will be followed by some of this. Then
we will venture into this non-linear conserve. So, after LQR this which is largely kind of
valid for linear systems, we will go and see how this, this, kind of simple concepts can
we extend to non-linear problems as well and that will lead us to this something call state
depending record equation approach, and then, there is a little bit further development
on that idea which is called theta d design. So, those things will talk. Then we will quickly
enter into this discrete time optimal control theory and then discrete time LQR and very
quick over view sort of thing, and that, then we will follow to a very different approach.
This is called dynamic programming and that will essentially lead to something called
Hamilton Jacobi bellman theory very famous. And then, we will see the, of that actually.
So, essentially the H J B theory over comes this curves of complexity, but essentially
it leads to course of dimensional so that those details will see that when we talk about
dynamic programming, and because of this curves of dimensionality, people are thought about
something called approximate dynamic programming in discrete settings. So, we will also discuss
that, and then, use this approximate dynamic programming concepts for adapt something called
adaptive creative design. And that and that will be a little subset
of that is something called single network adaptive critic design and that will be part
of our discussion next actually. Then we will venture into a different class of problem,
I mean different cross of formulation which is called model private to static programming
design. And again, this particular design is huge potential for guidance applications,
and we will see some of these examples actually there. Then after that, we will follow to
this estimation part of the story and these are all complementary problems - estimation
and optimal control sort of thing. So, we will see first what is called linear quadratic
observer.
So, that is from there, we will proceed to state estimation using Kalman filter, and
then, there is an idea of robust control through the concepts of optimal control and state
estimation. Once we put them in together, it leads to something like robust control
approach and all that we will see about that. Then on the way, we cannot forget that there
is a huge branch of optimal control that is of relevance which is called constrained optimal
control; that means, whatever solution we want it, it need not be only constrained wise
state to dynamics or system dynamics, but they can also contain some of these inequality
constraints of control and state; that means your control for example can be bounded between
certain values. So, those bounds can be incorporated here as well as the state bounce as well actually.
So, these are the essentially some of these inequality constraints.
And that, within that formulation, how do solve this optimal control problems. So, it
is possible to do and the some of this concept will see. Towards the end of the course, we
will go to every again at a very different approach. This is called transcription method,
and this transcription method is a very direct approach of solving optimal control problem.
That is what I just talk that you convert this optimal control problems dynamic problem
to a static optimization problem with large variables, large number of variables.
So, that is what is a, the problem get transcripted to a kind of a static optimization problem
and that is why it is called transcription method. Some overview of that, followed by
some little bit over view of what is called as Pseudo Spectral Transcription. So, that
is one of the things that is getting lot of attention these days and also is a very fast
approach in a way, so much fast that it can be probably we use online actually.
So, those kind of concepts can be discussed towards end of the course, and if time permits,
I will, I will also give a climbs of the similar concepts, that is, optimal control systems
is applicable to distributed parameter system, and distributed parameter systems are nothing
but systems governed by partial differential equations lot of application.
There are lot of application again this flow control and then temperature control and flex
flexible body control think like that is all fall under the distributed parameter system,
whether system dynamics is govern by PDS or partial differential equation which is a not
very common in OD domain basically. So, how do we solve or how do you take those problems
into account that, that is the topic of discussing next actually. So, this is how the course
of structured and then we very slight variation here and there, but largely we will do we
will stick to this, this, outline actually.
Some of references on the way you can, you can, think of probably buying some of these
books. Some of them are still available; especially the first one is very much available and it
is available in Indian edition as well. So, it is extremely cheap, and one of my favorite
systems or favorite books for linear system optimal control is the second one is very
clearly written and I like it. You can think of buying that, and especially the third one
happens to be a very classic book. It, if we remember, if we can notice, it is published
in 75, but it is still on print and it is per as the one book which is the most widely
referred ever in the in the entire control system domain basically not necessary only
in optimal control. So, that is that kind of implied book. It
is, it is never seen a second addition, but still the first addition contains enough material
for everybody to kind of look at it and then still keep on learning from it actually. I
strongly recommend the third one if somebody wants to get into these optimal control theory
and applications as well. Then there are other book, which are again there is a Dover publications
very cheap book from Stengel, talks about optimal control as well as estimation concepts.
And this, this, AB sage and C white book also extremely well written book. In fact, AB sage
and book that was one of the very popular book actually and it sustains its legacy as
well. In fact, this distributed parameter control and some of these sage equation concepts
and all I will take it from C. White actually.
So, there are other books well which is this one is another very classic book, lot of numerical
techniques are discussed very well here. So, some of those concepts can be read in detail
there and there are other publications as well. For example, these JL Betts it talks
about this transcription method very well; he is a very pioneer on that, and then, he
published many things on transcription method as well and that is what the book talks about
actually. Then there are other estimation theory concepts
and this is my most favorite book is Crassidis and Junkins very rigorously written, very
much well return as well and those of you want to buy some estimation books surely you
can by the this as well as d Simons book, and this is also extremely well written; however,
rigorous part is much more in Crassidis, and I also think there is second edition coming
up. It might already come up in late two thousand twelve or something that is my actually.
And obviously, there will be topics taken from literature as well; that mean some journal
and conference publications are also included as part of this course actually. So, take
a close note of all this things and ensure everything will be quite enjoyable experience
actually. So, let us go to some, some, degree of introduction and motivation first before
we go into this math, math, details and try to explore more.
So, come to the introduction and motivation first very basic concepts. So, first thing
is what is called a system variable. Then we talk about two type of variables - one
is I mean largely it can be classified something like input variable, output variable and kind
of state variables. So, what is input? Input can again be like some control input which
is a nothing, but a manipulative set of variables. That you can or we can rather vary as we wish,
and then, this can be computed precisely as well, but there are other class of inputs
and remember these are actually inputs to the system which is called noise input. In
other words, they, they, do come in input the system and they do disturb the system.
In other words, for example, if you go on a car, then the road condition and all will
affect the car drive. That is not a control system, but that is a noise input to the system.
Similarly, when aircraft flies, then there is a something called atmospheric dust phenomenon
and all. So, that will disturb the vehicle motion as well. So, those are noise inputs.
Usually they are non manipulative not in our control, but we have to live with that; we
have to handle that and that is how this necessity of robust control comes and all.
Then there are two types of output variables as well, and one is a one set of output variable
that you can measure by sensors. That is called sensor output, but there need not be something
that you want control. The control can be something same or different and the variable
that you really want to control the system or you want to have a performance measure
performing whether the system is performing well are not. Those things will be describe
something called performance output actually. So, sensor and performance output are may
or may not be same. Then there is something called state variables and the state variables
are nothing but a set of variables that describe a system completely; that means you will see
some examples, I mean you might have already seen in lot of text books early in system
theory. State variables are, there are certain restrictions. They cannot be, you cannot take
more state variables are necessary. You cannot take less state variables than necessary either.
There are again problems and advantages of that and like that, those of you are interested
can see a classic text book on these concepts actually.
And when we talk about system study in general and there are three category of system study
- one is model development which you can also include parameter identification in general
and typically will not talk about that. In other words, we will assume that the model
is typically available to us as far as control system synthesis problem is concerned, and
other words as a control system designer will not worry too much on modern development issues
and all that actually. So, we will assume that it is, it is, known to us in a away.
Then there are system analysis concepts; that means once you have a system model or a mathematical
set of equations for representing the model behavior, then essentially you want to know
whether system is behaving properly or not so that, that will lead to something called
system analysis. The typical example is whether the system is stable or not stable, whether
the system is on controllable or not controllable. So, those kind of issues are some part of
system analysis problem, and if the system analysis output is good, that means system
is behaving well; we do not want to do anything, but if something that needs to be done, then
we will go to the system synthesis part; that means, we design a set of control input in
such a way that the system performance is satisfactory actually.
In other words, we force the system to behave as we would like to see actually. That is
all, and typically will kind of give emphasis on the this type of issues and especially
on the system synthesis part of it, and analysis will be there in a weak sense, but largely
we will talk about system synthesis; otherwise how do we actually design a control variable
using these concept of optimization or dynamic optimization. That is what the main objective
of this course actually.
Then, we need, there are various classification again as a, when you talk about system, we
can classify in variety of ways - one is lumped parameter systems, where there is no relative
motion between two molecules of the system, and in other words, everything if the center
of gravity moves, then every other particle moves in the same way, same direction, same
velocity like that. So, that is a lumped parameter system, and then, there is a distributed parameter
system where we have relative, I mean the relative motion between two molecules are
also there as a along with the whole system getting. So, those are the system that can
be described only using partial differential equations, whereas this follows lumped parameter
system can be used in, I mean can be described in using this ordinary differential equation
and all that actually. Then, there are different classes of systems.
When you talk about continues times systems, time variable is treated as continuous variable
versus there is a discrete time system, where we talk about only discrete weak points, and
then, see the way read those weak points. Then there are ideas of quantized system,
non quantized system think like that, and there are something called quantized means
the variables are given at discrete number. The quantity of the variable can change only
in discrete numbers actually. Non-quantized is a something that varies continuously. In
other words, all rational numbers are also accounted for numbers actually, I mean, then
there is a concept called time invariant system and there is something called time varying
system basically; so that means, if the parameter of the systems can change in time, then this
leads to time varying system. For typical example is when, when, an aircraft moves in
a fuel and rocket moves in one fuel, the system is see’s the different mass as time goes
actually; so that means mass keeps on changing, and if that does not happened, then we talk
about something like time invariant system. You can think of, roughly you can think of
it is a kind of applying under in electric propels. If it is electrical propels, there
is no expendable mass goes there; it is only the charge and all that way.
But then, every other thing remains constant. There is no person moving there. So, they
do not move around; the mass remains same; moment of inertia remains same, and I mean
this kind of systems can be thought about something like time invariant system actually.
Then there are something like linear system or and verses non-linear systems, and largely
linear systems turns out to be linearised systems of non-linear system because nature
typically gives a non-linear systems. Anyway, coming back if you somebody analyzes
the system using discrete time theory as well as quantization theory, then it essentially
leads to something called digital control, whereas continuous time and non quantized
variable combination leads to analogue control, and it is so happens most of the time that
we confine in courses to the first line; that means we discuss lumped parameter system largely
in continuous system domain using non quantized variables assuming time variable systems and
linearizing the system, whereas the reality is all in the bottom actually mostly in the
bottom rather. So, the reality most of the systems theoretically
speaking at least can we describe in a distributed parameter system. Even though largely lumped
systems are also from the engineering point of prospective poin
…
Typically, it will be covering around aerospace engineering problems and applications thing
like that, but largely the techniques, I mean tricks and techniques will be very generic.
So, anybody from any other discipline can also use this concept at equal level. I am
Radhakant Padhi working as a associate professor in Department of Aerospace Engineering, Indian
Institute of Science, Bangalore.
So, let us see some of this introduction, motivation and things like that, before we
go to the details of the course. This is the main objective of this particular lecture
actually. So, before we start begin the some points to note. The topics covered in this
course are very generic with preferential emphasis on aerospace applications obviously,
because largely if we see this optimum control applications have been given by aerospace
engineering tragic optimum problems conventionally. And we also want to kind of develop the course
and this course is meant to equip the students with applied optimal control system and state
estimation concepts. So, as part of it, we will also take this missile guidance and other
equal missile guidance problems as well, which is nothing but an application of optimal control
theory, and from that prospective, we will discuss actually.
And the topics covered as I told are fairly mathematical; that means somebody who loves
mathematics can, can, enjoy it very well, but if you, if you are reasonably good in
mathematics, I think you will be able to follow it well, but that is a necessary requirement
of this particular course. And I also suggest all of you that, if you that as the course
goes along, you take a set of paper, pencil and all use start kind of practicing everything
yourself, then only the geometry comes here more actually.
So, the topics to be covered in this particular course is a first thing is a some review of
necessary fundamentals; that means we will talk about some numerical methods, matrix
theoretical, matrix algebra and whatever necessary background material, that, that, will be quickly
review. Then followed by static optimization concepts and that is a kind of a backbone
for optimal control in away, because even if we see one, one, method of solution of
optimal control is, is, first to convert it to a static optimization problem and then
take the advantage of static optimization and rotation all that. So, that we need to
know a little bit of a static optimization before we go ahead with the, with optimal
control principle, which is nothing but dynamic optimization.
So then, we will go to calculus of variation and see some of the concepts of. So, what
is this calculus of variation; obviously, it is a, it is a vast topic; we are not able
to cover everything, but whatever is relevant to us, I think that much we will we able to
overview before proceeding to calculus of variation application for optimal control
and that is what we will talk next optimal control through calculus of variation, and
then, that will be followed by some numerical techniques to solve this variation and calculus
problem, which essentially leads to these something call two point one revalue problem.
We will talk about that is the course develop actually.
But that is one of the reasons why it is computationally very kind of very challenging and computationally
over well and rather, and then, we will proceed for other techniques, reason techniques and,
and, classical as well as recent techniques to how to overcome some of those actually.
Anyway, after covering numerical techniques, we will move to this something called LQR
theory or linear quadratic regulator theory and that will be followed by some overview
of a flight dynamics, because most of the concepts you want to use in aerospace application.
So, it is better to know some of these flight dynamics on the way.
Then we will see how LQR is use for flight control applications. This is some of the
application like stability turbulent let us and then many variable announcement and think
like that. We will talk about that as an application of flow domain. Next their application LQR
will be optimal missile guidance and that is what lot of people do using linear system,
system, dynamics rather engagement dynamics between vehicle and target. That is what they
do. The missile and target, if you take relative motion and all the using that as a constraint
and then formulate some sort of a minimum problem, and then, solve it using l. So, those
details will talk after that. That will be followed by some of this. Then
we will venture into this non-linear conserve. So, after LQR this which is largely kind of
valid for linear systems, we will go and see how this, this, kind of simple concepts can
we extend to non-linear problems as well and that will lead us to this something call state
depending record equation approach, and then, there is a little bit further development
on that idea which is called theta d design. So, those things will talk. Then we will quickly
enter into this discrete time optimal control theory and then discrete time LQR and very
quick over view sort of thing, and that, then we will follow to a very different approach.
This is called dynamic programming and that will essentially lead to something called
Hamilton Jacobi bellman theory very famous. And then, we will see the, of that actually.
So, essentially the H J B theory over comes this curves of complexity, but essentially
it leads to course of dimensional so that those details will see that when we talk about
dynamic programming, and because of this curves of dimensionality, people are thought about
something called approximate dynamic programming in discrete settings. So, we will also discuss
that, and then, use this approximate dynamic programming concepts for adapt something called
adaptive creative design. And that and that will be a little subset
of that is something called single network adaptive critic design and that will be part
of our discussion next actually. Then we will venture into a different class of problem,
I mean different cross of formulation which is called model private to static programming
design. And again, this particular design is huge potential for guidance applications,
and we will see some of these examples actually there. Then after that, we will follow to
this estimation part of the story and these are all complementary problems - estimation
and optimal control sort of thing. So, we will see first what is called linear quadratic
observer.
So, that is from there, we will proceed to state estimation using Kalman filter, and
then, there is an idea of robust control through the concepts of optimal control and state
estimation. Once we put them in together, it leads to something like robust control
approach and all that we will see about that. Then on the way, we cannot forget that there
is a huge branch of optimal control that is of relevance which is called constrained optimal
control; that means, whatever solution we want it, it need not be only constrained wise
state to dynamics or system dynamics, but they can also contain some of these inequality
constraints of control and state; that means your control for example can be bounded between
certain values. So, those bounds can be incorporated here as well as the state bounce as well actually.
So, these are the essentially some of these inequality constraints.
And that, within that formulation, how do solve this optimal control problems. So, it
is possible to do and the some of this concept will see. Towards the end of the course, we
will go to every again at a very different approach. This is called transcription method,
and this transcription method is a very direct approach of solving optimal control problem.
That is what I just talk that you convert this optimal control problems dynamic problem
to a static optimization problem with large variables, large number of variables.
So, that is what is a, the problem get transcripted to a kind of a static optimization problem
and that is why it is called transcription method. Some overview of that, followed by
some little bit over view of what is called as Pseudo Spectral Transcription. So, that
is one of the things that is getting lot of attention these days and also is a very fast
approach in a way, so much fast that it can be probably we use online actually.
So, those kind of concepts can be discussed towards end of the course, and if time permits,
I will, I will also give a climbs of the similar concepts, that is, optimal control systems
is applicable to distributed parameter system, and distributed parameter systems are nothing
but systems governed by partial differential equations lot of application.
There are lot of application again this flow control and then temperature control and flex
flexible body control think like that is all fall under the distributed parameter system,
whether system dynamics is govern by PDS or partial differential equation which is a not
very common in OD domain basically. So, how do we solve or how do you take those problems
into account that, that is the topic of discussing next actually. So, this is how the course
of structured and then we very slight variation here and there, but largely we will do we
will stick to this, this, outline actually.
Some of references on the way you can, you can, think of probably buying some of these
books. Some of them are still available; especially the first one is very much available and it
is available in Indian edition as well. So, it is extremely cheap, and one of my favorite
systems or favorite books for linear system optimal control is the second one is very
clearly written and I like it. You can think of buying that, and especially the third one
happens to be a very classic book. It, if we remember, if we can notice, it is published
in 75, but it is still on print and it is per as the one book which is the most widely
referred ever in the in the entire control system domain basically not necessary only
in optimal control. So, that is that kind of implied book. It
is, it is never seen a second addition, but still the first addition contains enough material
for everybody to kind of look at it and then still keep on learning from it actually. I
strongly recommend the third one if somebody wants to get into these optimal control theory
and applications as well. Then there are other book, which are again there is a Dover publications
very cheap book from Stengel, talks about optimal control as well as estimation concepts.
And this, this, AB sage and C white book also extremely well written book. In fact, AB sage
and book that was one of the very popular book actually and it sustains its legacy as
well. In fact, this distributed parameter control and some of these sage equation concepts
and all I will take it from C. White actually.
So, there are other books well which is this one is another very classic book, lot of numerical
techniques are discussed very well here. So, some of those concepts can be read in detail
there and there are other publications as well. For example, these JL Betts it talks
about this transcription method very well; he is a very pioneer on that, and then, he
published many things on transcription method as well and that is what the book talks about
actually. Then there are other estimation theory concepts
and this is my most favorite book is Crassidis and Junkins very rigorously written, very
much well return as well and those of you want to buy some estimation books surely you
can by the this as well as d Simons book, and this is also extremely well written; however,
rigorous part is much more in Crassidis, and I also think there is second edition coming
up. It might already come up in late two thousand twelve or something that is my actually.
And obviously, there will be topics taken from literature as well; that mean some journal
and conference publications are also included as part of this course actually. So, take
a close note of all this things and ensure everything will be quite enjoyable experience
actually. So, let us go to some, some, degree of introduction and motivation first before
we go into this math, math, details and try to explore more.
So, come to the introduction and motivation first very basic concepts. So, first thing
is what is called a system variable. Then we talk about two type of variables - one
is I mean largely it can be classified something like input variable, output variable and kind
of state variables. So, what is input? Input can again be like some control input which
is a nothing, but a manipulative set of variables. That you can or we can rather vary as we wish,
and then, this can be computed precisely as well, but there are other class of inputs
and remember these are actually inputs to the system which is called noise input. In
other words, they, they, do come in input the system and they do disturb the system.
In other words, for example, if you go on a car, then the road condition and all will
affect the car drive. That is not a control system, but that is a noise input to the system.
Similarly, when aircraft flies, then there is a something called atmospheric dust phenomenon
and all. So, that will disturb the vehicle motion as well. So, those are noise inputs.
Usually they are non manipulative not in our control, but we have to live with that; we
have to handle that and that is how this necessity of robust control comes and all.
Then there are two types of output variables as well, and one is a one set of output variable
that you can measure by sensors. That is called sensor output, but there need not be something
that you want control. The control can be something same or different and the variable
that you really want to control the system or you want to have a performance measure
performing whether the system is performing well are not. Those things will be describe
something called performance output actually. So, sensor and performance output are may
or may not be same. Then there is something called state variables and the state variables
are nothing but a set of variables that describe a system completely; that means you will see
some examples, I mean you might have already seen in lot of text books early in system
theory. State variables are, there are certain restrictions. They cannot be, you cannot take
more state variables are necessary. You cannot take less state variables than necessary either.
There are again problems and advantages of that and like that, those of you are interested
can see a classic text book on these concepts actually.
And when we talk about system study in general and there are three category of system study
- one is model development which you can also include parameter identification in general
and typically will not talk about that. In other words, we will assume that the model
is typically available to us as far as control system synthesis problem is concerned, and
other words as a control system designer will not worry too much on modern development issues
and all that actually. So, we will assume that it is, it is, known to us in a away.
Then there are system analysis concepts; that means once you have a system model or a mathematical
set of equations for representing the model behavior, then essentially you want to know
whether system is behaving properly or not so that, that will lead to something called
system analysis. The typical example is whether the system is stable or not stable, whether
the system is on controllable or not controllable. So, those kind of issues are some part of
system analysis problem, and if the system analysis output is good, that means system
is behaving well; we do not want to do anything, but if something that needs to be done, then
we will go to the system synthesis part; that means, we design a set of control input in
such a way that the system performance is satisfactory actually.
In other words, we force the system to behave as we would like to see actually. That is
all, and typically will kind of give emphasis on the this type of issues and especially
on the system synthesis part of it, and analysis will be there in a weak sense, but largely
we will talk about system synthesis; otherwise how do we actually design a control variable
using these concept of optimization or dynamic optimization. That is what the main objective
of this course actually.
Then, we need, there are various classification again as a, when you talk about system, we
can classify in variety of ways - one is lumped parameter systems, where there is no relative
motion between two molecules of the system, and in other words, everything if the center
of gravity moves, then every other particle moves in the same way, same direction, same
velocity like that. So, that is a lumped parameter system, and then, there is a distributed parameter
system where we have relative, I mean the relative motion between two molecules are
also there as a along with the whole system getting. So, those are the system that can
be described only using partial differential equations, whereas this follows lumped parameter
system can be used in, I mean can be described in using this ordinary differential equation
and all that actually. Then, there are different classes of systems.
When you talk about continues times systems, time variable is treated as continuous variable
versus there is a discrete time system, where we talk about only discrete weak points, and
then, see the way read those weak points. Then there are ideas of quantized system,
non quantized system think like that, and there are something called quantized means
the variables are given at discrete number. The quantity of the variable can change only
in discrete numbers actually. Non-quantized is a something that varies continuously. In
other words, all rational numbers are also accounted for numbers actually, I mean, then
there is a concept called time invariant system and there is something called time varying
system basically; so that means, if the parameter of the systems can change in time, then this
leads to time varying system. For typical example is when, when, an aircraft moves in
a fuel and rocket moves in one fuel, the system is see’s the different mass as time goes
actually; so that means mass keeps on changing, and if that does not happened, then we talk
about something like time invariant system. You can think of, roughly you can think of
it is a kind of applying under in electric propels. If it is electrical propels, there
is no expendable mass goes there; it is only the charge and all that way.
But then, every other thing remains constant. There is no person moving there. So, they
do not move around; the mass remains same; moment of inertia remains same, and I mean
this kind of systems can be thought about something like time invariant system actually.
Then there are something like linear system or and verses non-linear systems, and largely
linear systems turns out to be linearised systems of non-linear system because nature
typically gives a non-linear systems. Anyway, coming back if you somebody analyzes
the system using discrete time theory as well as quantization theory, then it essentially
leads to something called digital control, whereas continuous time and non quantized
variable combination leads to analogue control, and it is so happens most of the time that
we confine in courses to the first line; that means we discuss lumped parameter system largely
in continuous system domain using non quantized variables assuming time variable systems and
linearizing the system, whereas the reality is all in the bottom actually mostly in the
bottom rather. So, the reality most of the systems theoretically
speaking at least can we describe in a distributed parameter system. Even though largely lumped
systems are also from the engineering point of prospective poin
…