AI, Path Finding

[JP]

What path finding algorithm are you using anyway? djkstar?, wavelet?

It's just a breadth first search at the moment, it seems to be doing the job pretty well and not causing any performance issues to it's gonna stay like that for now. I can't recall the advantages and disadvantages of breadth first right now but i should look them over and consider how good it is. It was one of the main algorithms i used in my AI uni project where i also used Iterative Deepening Depth First (Depth First is just a stupid algorithm for this sort of thing) and a really bad greedy best first search which i used for the stupid AI as it was far from guaranteed to return a path all the time.

As i've said before i'm just getting the bare bones structure of the library together so that Christian can implement it with the broad features and then we can look at what algorithm changes/optimisations may be required after that, which won't actually break the API.

[Acki]

why not when my reinvented wheel suits my needs better ?!?!?

Well then it isn't a reinvented wheel, because you made a new 4D wheel.

you're right !!!
it's like with formula 1 races, each team uses the preinvented wheel but creates it new for their needs, like gum mixture, profile and such...

well, I for myself like to use the basic A* algorithm and then I add things like way points, ground speeds, one way paths and such...

I suggest evreybody who wants to use path finding to learn the A* basics, it's realy easy to understand and you'll get it in a view hours (or even in less than 1 h)...
the tutorial MasterGod mentioned is prety good for this...

well, everybody has to make this choice on his own...
if you want to use a prebuild lib and it suits your needs, why not using this lib...
so it's on you, DDd !!!

[JP]

It actually took me ages at Uni for the A* penny to drop.... I think all through my second year (when i was first taught it) i didn't really get it and then in the third year a different lecturer spent like 15 minutes going over it and i was like Ohhh!! I get it now!

[Darktib]

If you want a library for pathfinding, you can check IA::Astar ( http://forum.thegamecreators.com/?m=for ... 40&b=5&p=0

It only need the STL, works using pathmaps. And the v1.0 will be downloadable in about 1 or 2 month.

Ok, I make self-advertising...^^
The 1.0 will be usable in C++ and Dbp

[DDd]

ohh... never got email notification on this thread, so sorry for the late reply.

I forgot that irrlicht is "just" a 3d engine, due to all other components irrklang, etc... being available i was just wondering if anyone built a game engine/framework that provides a more complete set of feature with AI features implemented... Like dynamic generation of navigation mesh and collision hulls based on game level geometry, stuff like that making use of parts of the game geometry to feed into the AI datastructures. I was just letting my mind wonder a bit lol

In regards to AI middleware there is a bunch of it, but the good ones are mostly commercial. Aigamedev website has a lot of info on that

In regards to implementing vs reusing i prefer to reuse, specially complex and feature rich framework that have many hundreds of men hours put into it. Not the case of implementing A* that takes a couple of hours at most, more if you go thru all the possible optimizations.

This is why i like the open source development model, you can leverage the work that many others have done (stand in the shoulders of giants) but still modify if the need arises. Then again you get that in commercial model too you just have to pay more for a source license hehe

Actually i was not thinking about pathfinding at all when i created this thread it was more of an open question to know what the irr community has been doing in terms of AI. Looking forward to see what comes out of JPs efforts with irrAI