formalization of A-life, paramerter managers and my project

[Fawzi Roberto Mohamed]

Ciao,

WARNING: my english is horrible, but I hope you will understand me.

I am working on a project that I call ECAS whose attempt was to build a
framework to simulate evolution a artificial life.
I developed Ecas on Nextstep/Openstep using Objective-C as language,
not Swarm because I didn't know its existence, but as swarm is much more
developed than Ecas, I am thinking of porting it to Swarm.

Rather at the begining I decided that I wanted dynamic agents. This gave me
quite many problems, here is my attempt of solution (for the moment quite
memory hungry).

I tryed to represent any information in Ecas with objects that I called
Parametri. Parametri can be any object, the only thing is that (for me) they 
conceptually represent some kind of information.
Parametri are contained in an object (as an Agent) theat I call BaseObject.
To extract the Parametri from the BaseObject you use Keys(objects like strings
or idnr) to identify them.

It is clear that this system standarize how information (Parametri) can be 
accessed, but such a system isn't of great use if you don't have a mean to 
distinguish the various Parametri; to do different things with different
Parametri.

I classify the Parametri using Types (objects like Keys, for example strings
or numbers). A Key can have many Types, for example the Key "E1" could have
Types "Energy" and "Number". A system (the KT System) keep track of the
Key-Types relations and let you make searches. Searches are very fast because
the KT System uses hash tables and can be performed on both Types and Keys
(Types and Keys are dual).

A parametro can be a BaseObject and this make the whole structure recursive.
The KT System also make a distiction between static and dynamic types, and
some other things that I won't discuss here, for them you will have the wait
until I publish the documentation of Ecas.

The KT System let one have dynamic contents in an Agent. To let an Agent also
have dynamic methods I have found two ways:
- adding Parametri with a special type ("active" for example) that respond to
  them and then either have a method of the BaseObject call them, using
  message forwarding or making the object that want call the method search
  these special Parametri in the BaseObject.
- making a method behaviour depend on some Parametri that can be modified

I think that to make ABMs really intresting mathematical models like linear
algebra and algebra in general can be very useful (I have seen an intresting
model of conjunction on Axiom done with non associative algebra), but
implementing everything with only one model is (at least I think so, but I am
not an expert) too restrictive, something more general to embed these models
is needed, maybe something like what I have described (at least I hope so).

After a so long message (have you seen that the last point in the subject is
the more developed? maybe I have written them in the wrong sequence), still a
little more on my project: I should publish something on internet at the SCSC 
(www.scsc.ch), I am ready , they have said they want do publish it, but it need 
some time to have an account...
I think that I will try to port the KT System to Swarm, (even if cof,cof I
still haven't installed it because I developed Ecas on Openstep, and Swarm
work on Solaris), so I will like some feedback (I will mail the adress of
the documentation of Ecas to swarm-modelling as soon as I publish it).

Regards,
Fawzi

Fawzi Mohamed
Regensbergstr. 80
8050 Zurich Switzerland
address@hidden