What is a good format for a 3d topview map?

Question

I am building a 3d topview game like GTA2. In this game the ground is mostly one level (except for tunnels and highways). Most of the ground is also city so the ground is usually covered by either roads or buildings. The buildings are simple 3d models which the player can walk around on. Most of the gameplay is 2d.

How should I model the map? I have considered the following options:

• Bitmaps, I think this gives problems when I want to add data smaller than a tile. Also the problem of bridges and tunnels seems hard to solve with this.

• Polygons, define all roads, terrain types, buildings with polygons and save that along with physics and texture information.

What approach would you take or have you ever implemented this or know of an implementation?

Answer 1

Are you going to use a terrain system ? or is it just flat ?

Bridges and tunnels do add complexity. To have bitmaps represent bridges and tunnels you could perhaps use a 2nd bitmap layer or even one of the channels to represent a second ground layer. (i just noticed that zacharmarz mentioned these considerations) I'm not sure, as there is always a way to make something work, but I'd imagine any commercial game using such sophisticated geometry would utilize a 3d engine. So unless you're an experienced game developer, I'd ditch the bridges and tunnels idea. At least until you've got some other more basic requirements met. Oh, unless those bridges and tunnels are part of the only traversable layer, then it's easier

You can add texture data onto the ground smaller than a tile. That's not a problem. And you can have objects (other than buildings) that are smaller than a tile, because any objects including buildings will be geometry which you have defined. All this talk is speculative though, because the solution options you have are numerous. It would be totally possible to build a bitmap based map system as long as you are clever (and patient) enough to overcome all the challenges. Indeed a lot of game terrains (2d & 3d) are built using heightmap's blendmap's ect which are simply bitmaps. BUT, if you want the power to be able to drop an object onto the ground, anywhere, it makes more sense to me to have in addition to the bitmap (defining ground textures / buildings or whatever) a collection of objects which have location properties : float x, float y, float z or Vector3 position. And run through that collection as well, rendering at their positions, just as you do for your avatar/player character.

Answer 2

I would prefer bitmaps, which will hold information about objects (buildings, roads, water, etc.).

If you have problem with tile / pixel size, you can separate "city / map" into several big tiles and each tile will have separate bitmap.

Problems with tunels and bridges can be solved by using more than one bitmap for tile / map. Or by using different color components. When you will be rendering you world, you for example just look, if there is no bridge over water, you are just rendering.

Answer 3

You can also do a hybrid approach: 3d with modular pieces. Make a set of 3d "tiles" and piece them together in a level editor. That gives you the fast editing speed of bitmaps but gives you the flexibility to add bridges and tunnels and custom set pieces.

Neverwinter Nights uses a system like this, if you're looking for inspiration.