Thanks for the quick response.
> Also we have heuristics to handle unit to "job" matching.
> Where job can mean: suppling, training, eating, getting cured,
> fighting, observing or clearing.
Well actually that's what I'm hoping to improve, ie allocating the most appropriate tasks to workers and allocating the most appropriate workers to tasks, as well as choosing the easiest way to perform the task (like gathering the most accessible resource for the task). I was hoping to implement a somewhat mathematical approach to do it, though I'm not sure about how the current heuristics work.
However I have seen workers being allocated for tasks half a map away when there's another task they can do right next to them, and sometimes they travel a fair distance away to get a particular resource for the building which was right next to them at first, even when the required resource is available from more nearby sources. And how about when a worker is already doing a task, with the resource already being carried, but suddenly decides to abandon task and go pick up some other resource even though the original task is still incomplete and easier to perform? I've seen these happen quite a few times, but I'm not 100% certain that they're not how things are supposed to be or just not what they seemed to be, so please correct me if I'm wrong.
Also, is there any documentation I can have access to which explains the general architecture of the game and detailed descriptions of the algorithms/heuristics? Or are these documented using in-line comments? I think I should read up on those before making any further assumptions...
Thanks.
On Mon, Jul 21, 2008 at 2:51 AM, Kai Antweiler <
address@hidden> wrote:
> in particular the micromanagement AI. I believe it
> can and should be improved because as the micromanagement is supposed to be
> taken over by AI, the AI that handles that aspect of the game should
> be completely reliable in producing optimal micromanagement, which is
> currently not the case.
I don't think that we currently have a specific AI to handle
micromanagement. Also optimal micromangement will not be
possible.
We don't have an ordinary rts with an AI on top that handles
micromanagement - if that is what you thought.
Globulation 2 is designed bottom up to solve micromanagement.
Currently we have optimal solutions to some subproblems (e.g: shortest
path). The shortest path solutions are heuristically enhanced to
population level by mechanisms that try to prevent collisions. And now
we heuristically try to reduce traffic jams for specific situations.
(Btw.: we still don't have an efficient three point pathfinding solution.)
Also we have heuristics to handle unit to "job" matching.
Where job can mean: suppling, training, eating, getting cured,
fighting, observing or clearing.
I guess the "micromanagement" code is scattered and disconnected.
Of course this could change ...
There are also unimplemented features (e.g. soughing or
transportation) that might require new micromanagement algorithms.
I also would like to have warriors move in group formation to their
targets taking a more or less appointed route.
So there are interesting topics concerning micromanagement that can be
worked on in glob2, but nothing with optimal solutions.
> If successful, and with the permission of the relevant parties, I wish to
> base my honours thesis on the work that I'd have contributed to glob2.
Sure, why not.
> If you have ever had
> any thoughts about this issue before, please feel welcome to share them with
> me :)
Done.
--
Kai Antweiler
_______________________________________________
glob2-devel mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/glob2-devel