RTS Map thoughts

Posted on October 7, 2012

I spent another shower session pondering how I can make an efficient 2D RTS work out, that is based on isometric tiles.

I’ve also been reading some of the source code for fheroes2 which gives me the perspective of how others developed a game they already knew all the specifications and requirements for, since there existed another implementation.

It’s not a real time strategy, but what I was most interested in was how they defined the map and possibly handled sprite animations.

I never got to the animations, but the tiles were interesting anyway. I also pondered how A* will fit in effectively. It seems it would be best to consider an on-demand heuristic with some sort of desired calculation length and a means of memory such that we don’t get put into a loop if things don’t work out for the heuristic approaches. However, I have to consider that in the memory, there will be a Fuzzy flag, meaning that the position previously considered taken, or bad, might be available again.

Such is the case if you have multiple objects moving in a cluster, the unit might be surrounded, and although they as a group are moving towards a similar destination, the unit may be surrounded and the only way forward is the spaces that others free.

Regarding maps, the idea I have is that there will be a quadtree of sectors, such that locations globally can be found in logarithmic time while possibly being sparse and expandable.

These sectors will be the size of 8x8 tiles. Regarding spacial qualities I’m considering that a tile might have three modes of spacial reservation.

  1. Entire Tile
  2. Tile is split into 4 subspaces
  3. Tile is split into 9 subspaces

Entire Tiles for example would be for a large mobile unit, or a tower or other structure.

4 subspaces would be for when you have units moving together and to ease movement of small units in groups.

9 subspaces would be when you have a central resource, such as a tree, where it takes up a central space, and spaces around can be considered as 9ths or 4ths when it comes to harvesting.

However, the 9 subspaces might just be considered 4 subspaces but not possible to take up the entire space with a unit.

What have I done so far?

I’ve played around and I’ve made a means to draw, and drag a map, so I’m beginning to understand the means of Input handling with Slick2D. I rather like it, and my own might follow it’s model.

I’ve made classes for Animation Data, Animation Sources, Animation Instances, so now I can grab a sprite sheet, which Slick2D can load I can pass it to an Animation Source, define some data, and the source can keep track of it, then for a given unit I can tick the Animation Instance, so that it updates frame by frame, and have a Volatile Image selected and drawn at runtime.

I specify Volatile because if it’s a dynamically generated Image, meant only for that frame, and to be destroyed after rendering.

So I also made a Volatile Image wrapper, which Inherits Slick2D’s Image class.

What to do next?

Get a Visual Unit class going, which has the ability for different directions if it is a mobile Unit.

Get a controllable unit class of sorts existing, so that I can give commands to a visual unit to move somewhere, change it’s animation states automatically, and so forth, which will be important to actually making a game unit come alive.


I’m liking my Java experience right now considerably better than I had over a year ago. My only complain tis that Eclipse generates over a gigabyte of garbage every 5 minutes! And When you use 16GB of RAM, and your Operating System OS X sucks at global memory management, you keep having to run purge so that it flushes out all that crap.

And this is with Eclipse minimized by the way. This same kind of activity is what really turned me off from Net Beans.