Re: [Repast-developer] Help! How can I use Repast for my multi-agent model

[Robert N. <rn...@gm...>]

Yiwen,

from what you have said, it seems to me that Repast would work fine
for your needs.  However, you will need to construct some of your own
framework that goes beyond what Repast supplies.  Most importantly it
seems would be the community where agents can leave and read messages.
 Doug Hofstadter had a model where agents went to a common place (he
called it a blackboard or workspace or something like that) and read
other messages and wrote down new ones message.  Although he was
working on a different problem, the concept might still prove useful.

- Robert





On 6/27/05, Zhang, Yiwen <yiwen@u.arizona.edu> wrote:
> =20
> Hi, Dear all=20
>  =20
> I have a multi-agent modeling of online community. Finding the right
> software has been torturing me for a while. I happened to find Repast las=
t
> week, and I was wondering whether Rapast is suitable for my model. I woul=
d
> appreciate it a lot if any of you can give me some ideas of the viability=
.=20
>  =20
> Basically, in my model, I have agents and public place.=20
> ---------------------------------------=20
> Agents:=20
> 1) each participant of the online community is modeled as an agent, and i=
t
> has a lot of parameters describing their learning capabilities,  social
> capabilities, communities's attactiveness to the agent, etc. These
> parameters are randomly assigned when the agents are generated, and will
> dynamically change over the time.=20
> 2) Agents are not always in the community. Each agent periodically visits
> the community once it is generated(based on some time distribution), and
> when the attractivenss of the community to the agent decrease to zero, it
> will never come back again (it's like it died)=20
> 3) every time an agent visit the community, agent will initiate message,
> read others' message, and reply others' message.The detail of the
> activities, such as how many message to generate, which message to reply,
> are controlled by the learning and social parameters of the agent.=20
>  =20
> Public place:=20
> 1) The online community is simplified as a public place, where the agents
> can post messages, read messages, and reply others' messages. Each messag=
e
> is represented by a set of parameters determined by the parameters of the
> agent who posted the message. Public place is represented by a list of
> messages. Messages are not necessary stored forever, and they will be
> cleared up after a period of time.=20
> 2) The interactions among agents are the achieved through the public plac=
e.
> This means that there is no direct agent-to-agent interaction. When agent=
s
> post messages, it potentially will affect other agents if other agents re=
ad
> the message and even respond to them.=20
>  =20
> --------------------------------------------------------=20
>  =20
> I have a set of interesting questions to answer based on the simulation
> results. For example, I want to study how the different composition of
> population (different types of agents with different learning capabilitie=
s
> and social capability) lead to the different growth pattern of the online
> community.=20
>  =20
>  =20
> Do you think Repast is good for my purpose? If it is, could you please
> provide me some suggestions on which components or steps I should take wh=
en
> implementing my model in Repast. If it is not, do you have any suggestion=
s
> on other software?=20
>  =20
> Thanks a ton!=20
>  =20
> Yiwen =20
>  =20
>  =20
>  =20
>