MASON-INTEREST-L Archives

October 2015

MASON-INTEREST-L@LISTSERV.GMU.EDU

Options: Use Monospaced Font
Show HTML Part by Default
Condense Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Sender:
MASON Multiagent Simulation Toolkit <[log in to unmask]>
Date:
Sun, 25 Oct 2015 07:39:49 +0000
Reply-To:
MASON Multiagent Simulation Toolkit <[log in to unmask]>
Subject:
MIME-Version:
1.0
Message-ID:
In-Reply-To:
Content-Type:
multipart/alternative; boundary=001a11c385a0a5c5050522e8f35c
From:
Ernesto Carrella <[log in to unmask]>
Parts/Attachments:
text/plain (2799 bytes) , text/html (5 kB)
It seems like using elements() to bag agents and then step them in a for
loop is a recipe for artifacts: agents in the top left corner always act
before agents in the bottom right.

I usually do what Russell does, keeping a reference to agents in a separate
collection and stepping that, but I prefer an ArrayList to aBag; partially
because of generics but mostly because I can call
Collections.shuffle(agentList,random) and randomize their order before
stepping them. The only caveat is that MersenneTwisterFast is not a kosher
Random object so you can’t pass it directly, but you can just pass a seed
from the MASON randomizer to a standard Java one (or extend the MASON one
so that it fulfills Randominterface).

As far as I could tell, shuffling and stepping is a bit faster than
schedule all your agents separately and letting the scheduler randomize for
you. But that might have been a fluke and I never tested this carefully
enough.

On Sun, Oct 25, 2015 at 4:13 AM Russell Thomas <
[log in to unmask]> wrote:

> I have accomplished this by having an łagents" Bag variable in the main
> class (SimState), and then every time I create a new object, I add it to
> the łagents" bag.  Then I iterate over all agents simply by looping using
> get() method.  If your agents ever die, then you also need to remove them
> from the łagents˛ bag.  This, of course, requires passing the SimState
> object to what ever object is iterating over all agents.  But, in my case,
> these objects involved UI, data collection, debug, and other
> administrative functions, and not objects within the model Iąm simulating.
>
> With this method, you never need to be concerned with where they might
> reside in any other data structure, including ObjectGrid or SparseGrid.
>
> Russ
>
> On 10/24/15, 9:56 PM, "MASON Multiagent Simulation Toolkit on behalf of
> Sean Luke" <[log in to unmask] on behalf of
> [log in to unmask]> wrote:
>
> >Iterating over all the objects in an ObjectGrid2D means iterating over
> >every single
> >cell in the array.  If you don't have a lot of objects relative to the
> >cells, this can
> >be very slow, and it would make more sense to use a SparseGrid2D instead.
> >
> >You could use a double for-loop to go through all the objects.  But
> >elements() isn't too slow, and it's O(1).
> >
> >Sean
> >
> >On Oct 24, 2015, at 10:37 AM, Axel Kowald <[log in to unmask]> wrote:
> >
> >> Hello Together,
> >>
> >> I just wonder what the best way is to iterate over all the agent on my
> >> ObjectGrid2D ?
> >> Currently I'm extracting a Bag with all agents by using the elements()
> >> method. It works, but I'm not sure how fast/slow it is. Would it be
> >> worth to maintain my own Bag and do the bookkeeping by
> >> inserting/removing agents when needed?
> >>
> >> thanx,
> >>  Axel
>


ATOM RSS1 RSS2