Re: New pheromone demo

út 20. 6. 2023 v 9:36 odesílatel Dave Raggett <dsr@w3.org> napsal:

>
>
> On 19 Jun 2023, at 23:43, Melvin Carvalho <melvincarvalho@gmail.com>
> wrote:
>
>
>
> po 19. 6. 2023 v 13:14 odesílatel Dave Raggett <dsr@w3.org> napsal:
>
>> Swarms can be used as a model for how web agents can work together to
>> achieve common goals.
>>
>> I’ve developed a new demo that shows ants scuttling about looking for
>> food guided by the pheromones they collectively deposit:
>>
>> https://www.w3.org/Data/demos/foraging/
>> ...
>>
> I really like this!  I've been thinking about web scale agents like these
> ants.
>
> What I was wondering about is giving each ant a URI.  Or perhaps giving
> the ant class a URI which could then be run to instantiate an ant.
>
>
> You could then have them roaming the internet to find food, bring it back
> to the nest, leave trails for the other ants.  Maybe the ants also have
> energy and will die without some share of the food.
>
>
>
Dear Dave

Been thinking about these questions alot, and there's no easy answers, but
I'll try and add some thoughts


> You need to develop the concept further. What does the ant actually do
> when “roaming the internet” and what kind of map is involved?  How is the
> map developed as a result of the collective efforts of the ants?
>

Back to biological basics.  All life transforms energy from one form to
another.  Food is a source of energy.  Life uses food to replicate.  Energy
and food are finite.  Food is finite and scarce.  There is competition for
food.  Ants "find" food.  Ants "consume" food.  Life consumes energy
through movement, discovery, eating.  Life can appear to learn or be
intelligent.  From these, but not limited to, rough observations you can
make a local system, or a network (web) based system.

Thoughts:

In a web based system ants have a URI.
In a web based system food has a URI.
In a web based system food needs to be scarce, finite and represent energy
in the real world.
The concept of eating requires a record of who ate what first.  Easy in
local.  In network you have a race condition, and need a tie breaker.
Ants and food need a llifecycle.

So some of these things need to be modeled in a multi agent network (web)
system.

A lot of orchestration is needed.

The only way I can think of for energy to be given to networked ants is for
humans to get the ball rolling and put it there.  ie humans pay the ants
with energy either on an altruistic or quid pro quo basis.  I must have
missed something here, though


>
> For a rather different scenario, consider a set of mobile agents that use
> Lidar or video to collectively map their 3D environment as they move
> around. Lidar produces point clouds, which are amenable to neural gas
> algorithms. Video is  more challenging, as it involves segmentation, object
> recognition, frame to frame consistency and depth perception. Each agent
> performs local processing as well as exchanging messages with other agents
> to collectively develop a consistent overall map.
>

Yes, this is great.  I've not really experienced the world of physical
agents yet.  I've more been thinking of purely digital.  but there's a link.


>
> For what’s worth, the ant foraging demo could be readily extended to model
> the diminution of food sources as the ants carry small amounts away. New
> food sources could be made to appear from time to time, and even new
> obstacles. Pheromone evaporation is slow, but key to enabling the colony to
> adapt to changes in the environment. However, too fast an evaporation rate
> will preclude the colony from learning good routes. Note: in the demo, the
> ants that are “lost” have no trails as they been away too long.
>

Tweaking the environment I think will make huge differences to the
evolutions of the colonies or evolving ant species.

Sort of high level hand wavy thoughts, not sure how close this is to be
turned into a working design.  I must have missed loads of things here!


>
> Best regards,
>
> Dave Raggett <dsr@w3.org>
>
>
>
>