Re: Real Space

[Sven N. Thommesen]

At 01:10 PM 4/11/2001 -0500, you wrote:
> "Marcus G. Daniels" wrote:
> > One way to avoid having agents reason in absolute space is to have all
> > agents inherit from an abstract base class that does these collection
> > procedures in one place, e.g. a getNeighbor (radius), method.  That
> > way you can hide the coordinate system of the space from the derived
> > classes.
>
> "Darren Schreiber" wrote:
> "While I illustrate that with the canvas and graph, the search of the
> space is
> done by parsing through the lists of agents and is independent of the
> display issues."
>
> Suppose I wanted to create a Grid2dContinous class, with basically the
> same interface as Grid2d, except inside there I want to use a "better
> faster" collection object into which the objects are placed.  I want to
> more quickly find all agents within r units of (x,y).
> getNeighbor(radius) is terribly slow if it has to cycle through a whole
> list of agents (as Darren can probably testify).
> At least, we want to take some subset of the agents before checking if
> they are within r units.

If I remember correctly, what I did in my old SvenSpaces was to pre-load 
each cell with a list of pointers to cells within a given radius of the 
current cell. Only useful if the neighborhood size (vision length) is a 
fixed parameter for the simulation, obviously. And we trade memory for 
speed, as usual.

Sven


                 ==================================
  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.