Kalman Filter

Hello world, welcome to the
move 37 course from school of AI

Today we will learn

how to teach
the missions to learn

from a noisy
fused sensor data

What is Kuhlman filter?
Kalman filter is

an optimal estimation algorithm

Let's see how we can do this

We are going to use
two exciting examples today

The first example is
an engine of spacecraft

The second example is
how a self-driving car can

navigate it using the
different sensors that it contains

If you are not
familiar with the topic

You may be asking yourself
what is this filter?

Is it like a air filter or
even a coffee filter?

Let's see. Today
we will discuss how

we can use the Kalman filter
for various purposes

The first example
that we are going to see is

how we can measure

the combustion chamber
of your spacecraft engine

First we will see
how a Kalman filter

can be used to
estimate a system state

when it cannot be
measured directly

we cannot go
certainly go into

a combustion system
and see how hot it is

Let's go bit deeper

how the spacecraft engine is working

A spacecraft needs some
thrust to propel itself into the space

The thrust is provided by the fuel

The fuel should be burned
inside the combustion chamber

to a certain temperature
to give enough thrust

We need to monitor this
combustion temperature

to keep the thrust in a
certain level

that the engine needs to propel

Think of this as an acceleration

when we are pressing
our acceleration pedal

we are putting more fuel
so the car goes faster

The same thing happens here

This seems to be easy task, right?

Can't you place a
simple sensor to do this job?

Okay this is certainly
not an easy task

Let's see why

That's too hot right?
The sensor would probably melt

Instead of this

We need to find
a way to place the sensor

in a cooler surface
but it should be closer to the chamber

The chamber here defines
the combustion chamber

where the fuel is being
burned to give the thrust

Let's see how we can
solve this problem

This is a expanded view
of a spacecraft engine

Here this orange colored area is
the combustion chamber

This is where the fuel is being
burnt to create the thrust

We cannot have
a sensor inside but

we can have the sensor
as close as to this chamber

T in is the internal temperature
that we want

but it is not available
for measurement

T external is the
measurement that is available

In this situation
we can use the Kalman filter to

find the best estimate
that we can get of the internal

but we cannot measure
From out we can

This is a very good example
where we can use Kalman filter

So to the next example
this is my favorite one

Self-driving car

Who doesn't love
self-driving car, right?

Ok, so self-driving cars
have many sensors

that it is laying on to
drive safely in the road

We have GPS
We have lidar

We have radar
We have a video camera and

we have more sensors
and all of these are

controlled by
a central computer

To know the location
of the car, we can use GPS

But will that work always?

What about if we are inside a tunnel

or if you are in between
two long big towers

then we will get so many of
the GPS error

and we will lose the track of the car

The car will not know where it is
positioned in the globe

So to avoid this problem
we have lidar

light detection
and ranging

which will give
the current position and

we have radar

which will monitor the position of
the other vehicles nearby

From this,
we will be able to gain

the knowledge of
where we are going to go

if the road is free,
we will move forward

if it is not, we will wait
untill it moves

We can use a Kuhlman filter
to combine these three sensors

and get the exact location
of the vehicle

Thank you for watching this video

Please go to the following link
so you can read about

how the Kalman filter
is working and the math behind it

Thank you