Re: My project

[Vladimir Jojic]

On Mon, 19 May 1997, Mr X.Y.Huang wrote:

>  Following Vladimir's suggestion (see his/her message 

Hey! It's His ... from now on refer to me as He ... :) (ok, never mind,
this is the first time something like this happened to me. Hmh, How should
one respond to this ?) 

> The second step is to simulate 
> travellers' behaviour according to the transport model to test 
> whether the model is correct, or to see the aggregate performance of 
> travellers under this model.  We call the second step the 
> implementation modelling which is also my main task.  More 
> specifically, the implementation model is a tool for testing the 
> transport model(s).  We consider the implementation model should have 
> the following features: (1) accomodating most transport models (some 
> have been developped, some are been developing, and there might be 
> new transport models in future), (2) agent-based --- taking each 
> traveller or a group of homogeneous travellers as an agent, (3) 
> activity-based --- each agent will have a chain of activities (e.g., 
> geting up --- departure --- arriving work place --- working --- lunch 
> --- working ---- going home).   So, in the implementation model, each 
> agent should make various decisions at the right time: what 
> activities it will have for the day (trip plans), what is route 
> choice, what is the departure time, etc. As we said how to make the 
> decision is a matter of transport modelling, so when the agent should 
> make a decision, it will call  a suitable transport model which will 
> produce the required decision. As far as the transport model 
> researchers are concerned, the implementation model should also 
> consider the possibility of interaction between agents in future 
> transport models. 
> I would like to get help from you relating to what multi-agent 
> language I should choose. Swarm may be a candidate, but I do not know 
> whether it is suitable for my project? 

Nice thing about Swarm, is that you can model spaces, agents, anyway you
like. 

You could create a space representing a graph assigning each node the
prices per km (mile) and the rest of the needed stuff (somebody is already
working on this (graph library), so if they read swarm-support, they
will,hopefully, respond to this, with the state of the project). Further
on, you can create agents for each transportation mode (bus, cars ... even
walking (not really natural reasoning)).

Transportation models, could be specified in a goal-action chains (just
create lists, as you do in C, putting information about the next action,
conditions, what transportation mode to use, etc in the element of that
list...). You could create a linked list of transportation models ...

Create an agent called traveler, set its internal variable to point to a
node in a graph. If agent wants to use certain transportation model, it
will browse through the list of transportation models. After agent chooses
the one it wants to use, agent can go down the chain acting as
transportation model says.

Between two transportation model usage, agent could use the goal-action
chains for other daily routines, and incorporate transportation for
achieving some of them (i am hungry -> want to go to restaurant -> go to
restaurant -> chose t. model (and execute) -> order lobster). 

Grouping agent, is an interesting issue in any environment, swarm or not
(I will write about this in couple of days, in my abstract, there won't be
any jokes there, well maybe just one -- good one, for a change;). With the
swarm you can define the collection of agents, putting them under one
object, whose behavior you can control ...

Allowing agents to communicate, is the least of the problems ... 

Let me be clear, swarm does not support, goal action chains, I just
suggested possible solution (I think it is a good one, I could expand 
the idea in a separate, personal mail)....

I think that Swarm is good platform for you, because, it gives you freedom
to design your own environment, agents, and their interactions, with very
little constraints (if any). 

I have read about artificial life, a lot (especially lately), and I have
been playing with it for a long time (does anybody remember c/pascal/tcl-
-robots?:) and I have yet to see, any package that can match Swarm.

> > PS. Please, don't forget to post your abstract (project description) to
> > the swarm-modeling, so that people who are working on a similar project,
> > could get in contact with you (and you with them)... 

I was told so, and I like to pass on the knowledge :) ...

Regards,
Mr. Vladimir Jojic





                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.
                  ==================================