Re: [Swarm-Modelling] starting ant system

[Vitorino RAMOS]

Dear Nelson and Swarm modelling list colleagues:

Real ant colonies "avoid" somehow this tail following problem by having one 
osmotropotaxic sensitivity (a kind of instantaneous pheromonal gradient 
following), and on the other hand an sensory capacity which describes the 
fact that each ant's ability to sense pheromone decreases somewhat at high 
concentrations [2]. From here you can have an equation which measures the 
relative probabilities of one ant moving to a new cite. Also, a bias is 
introduced on the moving direction (by having implicit directional 
weights), in order to avoid U-turns which are improbable (but possible) in 
real ant colonies. The whole system however, is completely stochastic. That 
is, pheromone is a stimulus, not an order. Even if not probable, one ant 
having one neighborhood cite full of pheromone may "refuse" to visit it, 
and walk on different directions. This ability may sound strange, but can 
lead the whole colony to better adaptive behaviors.

Based on these real behaviors I have used a very simple 2D model which 
emerge paths, extending it to multi-dimensional environments (e.g. 3D) or 
for handling multi-dimensional objects [1 to 6]. These swarms may tackle 
objects (your food or obstacles) as images [3][1]  or words [4], coded in 
numerical vectors (or obstacles in the environment [2][6]), and are able 
among other things to produce clustering, classification and so on.

So my feeling is, if you chose the right parameters (check for instance the 
ones I have used), you can avoid the effect you are presenting us. There is 
no need at all to use different pheromones (i.e., different ant castes). 
However, if you feel the need of several castes (depends on your problem 
and aim), the model described here can again easily be used. It's only a 
matter of using two ant populations, with different pheromone sensitivity 
(A ants only smell pheromone A, B ants smell B), each governed by the same 
multi-dimensional or 2D swarm foraging model. Meanwhile, recent results 
point to high adaptive behavior on dynamic landscapes [2], and the ability 
of the swarm to self-regulate it's own population [6].

Hope it can be helpful. Please let us know once finished. Kind regards, 
Vitorino

~ v. ramos, http://alfa.ist.utl.pt/~cvrm/staff/vramos, "Interactions among 
many sporuliferous and ubiquitous abstractions may lead to increasing 
reality", V. Ramos, 2001.

----------------------------------------------------------------------------------
All links have the respective PDF paper available for download:

[1] "Artificial Ant Colonies in Digital Image Habitats - A Mass Behaviour 
Effect Study on Pattern Recognition",
 http://alfa.ist.utl.pt/~cvrm/staff/vramos/ref_29.html

[2] "Social Cognitive Maps, Swarm Perception and Distributed Search on 
Dynamic Landscapes",
http://alfa.ist.utl.pt/~cvrm/staff/vramos/ref_58.html

[3] "Self-Organized Data and Image Retrieval as a Consequence of 
Inter-Dynamic Synergistic Relationships in Artificial Ant Colonies",
http://alfa.ist.utl.pt/~cvrm/staff/vramos/ref_39.html

[4] "Self-Organized Stigmergic Document Maps: Environment as a Mechanism 
for Context Learning",
 http://alfa.ist.utl.pt/~cvrm/staff/vramos/ref_42.html

[5] "Web Usage Mining Using Artificial Ant Colony Clustering and Genetic 
Programming",
http://alfa.ist.utl.pt/~cvrm/staff/vramos/ref_48.html

[6] "Varying the Population Size of Artificial Foraging Swarms on Time 
Varying Landscapes",
http://alfa.ist.utl.pt/~cvrm/staff/vramos/ref_59.html

among many others available.
----------------------------------------------------------------------------------

At 18:37 14-03-2005, you wrote:
> Hello all,
>
> I'm starting the design/development of an ant system.
> I have one question, that i don't know if there you
> could have a better solution.
>
> The scenario of the test is a plan with X*Y positions.
> Each position could be: Food, water, or free terrain.
>
> My problem is that if an ant put pheromone at a given
> position I can have the possibility to put the ant
> running in cycle.
>
> To avoid this the ants put pheromones on each position
> with a different 'smell', based on ant genes.
>
> So the ant follow other ants pheromones, but avoid own
> 'smell'. Is this the normal solution?
>
>
> Nelson Faria
> +351 913495950