« Return to Thread: Model design: a single context or different contexts for each agent type?
Re: Model design: a single context or different contexts for each agent type?
by Jonathan Ozik-2
::
Rate this Message:
Chloe,
There are many ways to structure your model, but we generally recommend starting simple then adding complexity as the need arises. Given this, we'd recommend using one context to start.
Jonathan
On Jul 9, 2012, at 5:49 PM, Chloe Bracis wrote:
> Hi all-
>
> I'm curious how people design their models in terms of contexts and projections when there are multiple types of agents. Do you put everything into one context? Use multiple levels of sub contexts? When looking through the examples, many seemed to have one type of agent or use a single context typed Object (which seems to defeat the purpose of generics). I've read up on contexts and projections, but am still not sure the best way to proceed.
>
> A few more details about what I'm trying to do if it's useful. I have a 2D model with agents (currently one type, eventually several types) moving in continuous space, as well as a grid of stationary agents (multiple types) which change over time. My mobile agents interact with the stationary agents using a spatial kernel (i.e. a bivariate probability distribution), so I created my own subclass of ContextGrid to enable this interaction across the entire grid. If I have everything in one context (typed Agent which is an abstract base class for all my agents), I have a problem adding projections, since the type of my grid is a subclass rather than a superclass of Agent (why does the projection type have to be a superclass of the context type?). I've thought about breaking each agent type into it's own context, but I'm not sure if I will be able to display things together that way (e.g. the mobile agents overlaid over the stationary objects in the grid).
>
> So I'm looking for advice on how other people have structured their models that have been successful with multiple agent types.
>
> Thanks,
> Chloe
>
>
> ------------------------------------------------------------------------------
> Live Security Virtual Conference
> Exclusive live event will cover all the ways today's security and
> threat landscape has changed and how IT managers can respond. Discussions
> will include endpoint security, mobile security and the latest in malware
> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
> _______________________________________________
> Repast-interest mailing list
> Repast-interest@...
> https://lists.sourceforge.net/lists/listinfo/repast-interest
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Repast-interest mailing list
Repast-interest@...
https://lists.sourceforge.net/lists/listinfo/repast-interest
There are many ways to structure your model, but we generally recommend starting simple then adding complexity as the need arises. Given this, we'd recommend using one context to start.
Jonathan
On Jul 9, 2012, at 5:49 PM, Chloe Bracis wrote:
> Hi all-
>
> I'm curious how people design their models in terms of contexts and projections when there are multiple types of agents. Do you put everything into one context? Use multiple levels of sub contexts? When looking through the examples, many seemed to have one type of agent or use a single context typed Object (which seems to defeat the purpose of generics). I've read up on contexts and projections, but am still not sure the best way to proceed.
>
> A few more details about what I'm trying to do if it's useful. I have a 2D model with agents (currently one type, eventually several types) moving in continuous space, as well as a grid of stationary agents (multiple types) which change over time. My mobile agents interact with the stationary agents using a spatial kernel (i.e. a bivariate probability distribution), so I created my own subclass of ContextGrid to enable this interaction across the entire grid. If I have everything in one context (typed Agent which is an abstract base class for all my agents), I have a problem adding projections, since the type of my grid is a subclass rather than a superclass of Agent (why does the projection type have to be a superclass of the context type?). I've thought about breaking each agent type into it's own context, but I'm not sure if I will be able to display things together that way (e.g. the mobile agents overlaid over the stationary objects in the grid).
>
> So I'm looking for advice on how other people have structured their models that have been successful with multiple agent types.
>
> Thanks,
> Chloe
>
>
> ------------------------------------------------------------------------------
> Live Security Virtual Conference
> Exclusive live event will cover all the ways today's security and
> threat landscape has changed and how IT managers can respond. Discussions
> will include endpoint security, mobile security and the latest in malware
> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
> _______________________________________________
> Repast-interest mailing list
> Repast-interest@...
> https://lists.sourceforge.net/lists/listinfo/repast-interest
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Repast-interest mailing list
Repast-interest@...
https://lists.sourceforge.net/lists/listinfo/repast-interest
« Return to Thread: Model design: a single context or different contexts for each agent type?
| Free embeddable forum powered by Nabble | Forum Help |
