World Structure

G

Guest

Guest
Archived from groups: rec.games.roguelike.development (More info?)

This is something I found while searching on the depths of my devblog
here it goes:

So... while thinking in the basic AI modules for townspeople and
monsters, I stumbled uppon the need of defining the structure of the
locations.

Off course, the locations would be the in-detail parts of the
overworld, like the towns, the dungeons, the caves, the fortresses
etc...

I had tought on two alternatives:

* Making GWorld a container of "BigMapCell"s, which would describe the
overworld, and a collection of Locations, conformed of "MapCells".

* Restricting the World to a cube of MapCells, as was initially tought

The second option, while may seem bad, is very good in allowing the
Action performance model to work. There's only one space in which all
actor interact. However, it is cumbersome to travel long distances
here, and makes multistore buildings resource hungry to implement.

The first one, is very good for the physical model, but makes things
hard for the Actor Interaction, I was thinking in a good way to tackle
this; the overworld would not have Actors, MapCells or Items; but Units
and BigMapCells. So, the player would command the Units while in
overworld, going a bit Turn Based Strategy.

Everything is ok up to now... but the problem here is that the units
are composed of Actors too.... and the actors must also exist into
smaller scale locations. So the units and actors must be
synchronized... this makes things very difficult indeed....

So... how to handle this... I am still machinating... perhaps a mix of
both things, handling the actors independently and the units as an
abstract Actor that knows about all of the members of it, and may
influence the behaviour of their members....

Any thoughts???

--
Slash
 
G

Guest

Guest
Archived from groups: rec.games.roguelike.development (More info?)

At 21 Sep 2005 15:27:23 -0700,
SZDev - Slash wrote:

> Any thoughts???

How about just storing the important locations as nodes of a graph,
and then remembering the paths between those locations as the edges
(and the spatial description of the map, ie. a list of steps).

Then you don't need lots of memory, the graph can be generated on the
overworld map generation, and it's very fast for the agents, who just
pick the location they want to get to and then follow the saved path.

You can tweak the path following algorithm of the agents, randomizing it
a little bit (so that not all the agents use exactly the same route), or
adding some additional logic (like avoiding danger).

--
Radomir `The Sheep' Dopieralski @**@_
(TT) 3 Waaah!
. . . ..v.vVvVVvVvv.v.. .