MoveMaps

[Auntie Mangos]

I dont know, but DiSlord made some changes to map format here and here which needs to be added to movemaps, and I think G3D lib has been updated.

[Auntie Mangos]

the thing i was wondering about is if this tile limit is all tiles, or just tiles loaded into navmesh at some point. if its just the number of tiles that can be loaded at once, we can add more poligons per tile, just making sure we do not load to many tiles at once.

Detour creates the maximum number of tiles possible right from the start (in dtNavMesh::init)

It then uses those blank tiles as a pool, and that pool behaves like a stack backed by m_tiles[m_maxTiles]

m_nextFree is the stack pointer.

when you add a tile, Detour pops the stack and uses that tile

when you remove a tile, Detour clears the data in it, then pushes it on top of the stack

So we aren't dealing with a dynamic number of tiles. We're stuck with the 12bit tileID if we use 4096 tiles, unless we re-write all of the code that uses salt. Do we really need 10bits for salt? (question in the back of my mind is do we really need salt at all?)

edit: I don't think it would be a major re-write to reduce salt to 5 bits or something. Maybe just change one line in init!

And even if we didn't have these limitations, I'm sure there would be problems with not being able to load all tiles simultaneously.

Grids staying in memory permanently because of mangosd.conf GridUnload = 0

People may even have a lot of ram and prefer to pre-load all grids + tiles, as in -New-Map-Terrain-Management-System"][Map/Grid] New Map/Terrain Management System

[Auntie Mangos]

Ok, I have made some progress. Nothing new, mainly the abstraction part I mentioned.

Started new thread at http://getmangos.eu/community/showthread.php?13557-mmaps-redux, go there from now on.

[Derex]

Hi,

This is the current work on the mmaps project, which aimed to add pathfinding to mangos. If somebody wants to code it.

Me and Neo2003 took the code and fixed the all the bugs we found, ported it to autotools build. Made it perfectly generate all surfaces around the "world".

http://filebeam.com/a9d136a33168f6c1621535cc96beb993

What needs to be done is:

* There is missing a reliable way of finding out what are the points around a certain point.

* The actual logic of which paths will be finded is missing.

If somebody is interested in continuing the project I pls contact me for better explanation of the sources.

Screens (by Neo2003)

[Derex]

The red dots are the mmaps dots. The white lines that form the surface are made by navmeshgenerator. The blue is vmaps.

PS: The forum alows 4 images per post.

[EDIT] Looks like somebody hacked imageshack, I wont remove the pics in case they recover from backup or something.

[Cypherjb]

Out of interest how do you intend to implement the actual pathfinding?

A*? Dijikstra?

Also, will it be done on-the-fly or will paths be pre-generated based on a navmesh from the MPQs.

[Derex]

A* on the fly ... which means very slow and a lot of memory usage.

But inplementing accurate navmesh for all this surface maybe very complicated and may be not that accurate.

[Cypherjb]

A* on the fly ... which means very slow and a lot of memory usage.

But inplementing accurate navmesh for all this surface maybe very complicated and may be not that accurate.

I know a guy who is using a navmesh for his bot platform and its highly accurate. So it is very possible to get accurate navemesh generation. Also, resource usage wouldn't be overly high as long as the algorithm is implemented correctly and you're not constantly calculating huuuge paths.

[munche]

IMHO D* looks better then A* or Dijikstra

[Birdman]

Wow this looks amazing derex, great work as always

I have a question, can this be made into a git branch so it can be developed and handled easyer, and afterwards merged into the master branch ?

[Derex]

The problem is that there is no body to develop it

I dont have that much free time.

[Birdman]

I would hate for all of this good work to go to waste, but is there anybody who know how to work on this ?

If we can't find anybody this will go abandoned.

[DasBlub]

why you not share it with us on a git branch (at github) ?

so those who want/have time can fork it and try to expand it.

[alexluana]

wow this would be the ultimate patch for the server...1 step CLOSER to millstone 1 ...

Cheers and I hope someone will help out in this one ...

[Nekac1]

Interesting, i think.

[KlaSter86]

noone work on it ?

[XTElite1]

I certainly hope someone has contacted you by now. If I had anywhere near that talent level I would, but I don't.

[Derex]

no-body contacted

[alexluana]

maybe it's 2 hard for many ppl...

That may be why noone contact u...

That leaves only one solution.....u continue it sometime

Cheers

[Birdman]

I dboubt anyone can handle this patch and even if they can they either don't have the time, or mood to get into it.

Derex is there any way the we, the unexperienced progammers can help ?, maybe even do some chores around your house to save you some time (<<kidding) but seriously, this is very important to me, i will help in any way i can, just telll me how.

[duckman]

I'll wash your car!

[Derex]

Somebody contacted me.

[zhenya`]

very nice !)

[alexluana]

OMG ... give me a good news Derex...pls

Cheers

[rogermangos]

This will be the most amazing patch if committed.

If it implemented, it will make mangos infinitely close to offy one.

Any advancement? news?