News and opinion

Back to overview


What is the structure of a brain within an ANS? Part 2

Now that we have an embedded system, means the body hardware,
which consists of sensors, actors and internal organs,
we will have to connect this body to the brain, means the FPGA board,
where we will then develop the corresponding ANS software.

For this purpose, we have to know more about the brain itself.
As a brain is always dependent on the given body structure,
this implies that different skills have to be implemented,
according to the species related to this body structure.

As an example, let us take the locomotion skills of 3 species,
a fish, a mammal and a bird, they all can set their body in motion.
However, each of them will use with different body parts, which
will then be used by the brain, to execute the movement of the body.

A fish needs to swim, so the body has fish fins,
while the mammal has legs and the bird also has wings.
The brain needs to coordinate the adjacent muscles,
in order to perform the required movement(s).

But hey, the movement is different, as the surrounding environment,
therefore each brain will have different structures for this purpose.
Furthermore, at birth, some can move immediately, like the fish,
but the mammal and the bird need to learn first their skills.

This means, that not only the movement itself is different,
but also different stages of complexity are required.
Therefore the brain has to have different regions,
which can be seen, when analyzing the analog(real) brain.

We humans even have a very complex brain region for this, called cerebellum.
It is used so extensively, that this brain region is a separate brain.
And you know something else? In the first 3 years of our existence,
this is the location, where we "store" all our physical abilities.

Let us see, what is the structure of a human brain, in terms of neurons.
A human brain has about 86 billion neurons on average, for all regions.
Did you know that the cerebellum has 69 billions of neurons, so 66 %?
And our "superior" neocortex has "only" 16 billions neurons?

But this complex neocortex region is shared with other "simple" functions,
like perception and representation of "simple elements" for each sensor?
But then again, the sensor processing is also dependent on the higher brain functions,
which on their side require time, in order to learn, what we perceive with this sensor.

So, here we have it, the brain consists of different regions,
each with a dedicated size, structure and complexity.
Each region performs only a specific task, simple or complex,
which depends on the corresponding body, means species.

So, speaking in layman terms, a brain consists of 4 main ares:
First area is a sensor driver, which connects and converts the input sensor signals
into neuronal information, which is afterwards processed accordingly.
This processing depends also on the required complexity, both sensor and brain alike.

The second big area is the logic processing, which is triggered by the sensor driver.
In this area, the information is processed, means analyzed and stored.
As a result, there is new information generated, which will be send
to the corresponding muscles, to enable an appropriate body reaction.

The third big area of a brain is used to act, means the body reaction.
For this purpose, the brain receives from the processing area the corresponding information
and converts this into muscle signals, which afterwards will be the body reaction.
As a result, the body, means the embedded system, will do something in the real world.

The fourth big brain area is the one, which is responsible for the internal organs.
These organs, like stomach, heart, liver, etc. are also connected to the brain.
They send the "needs" of the body to the processing area, which will respond to it.
The stomach "says it is hungry" , so the logic processing sends the body to look for food.

But, a body has more than only one sensor, it has many sensors, of different types.
Each sensor has a corresponding process area, same as each limb and each organ has one.
Furthermore, each processing area, can have different levels of complexity,
which depends on the body structure, as sensors, actors and inernal organs, and skills.

So, now we need to establish, which and how many sensors need to be connected and how.
As the BNS has always only one type of interface, the ANS should also use only one.
In our case, I have made the decision to use only Ethernet, which has a great benefit:
it is bidirectional, means it can receive and send information simultaneously.

When creating different body structures with different sensors and actors,
then such a "universal" interface has many benefits, also for the future.
Why? Because the connectivity is always the same, we do not need to create new ones.
Besides this, the Ethernet interface is also used to connect many ANS together.

What about the self-learning mechanism within an ANS? This is the topic for next time.

Back to overview