Yo @Cedric … I saw that functionality in the demo as well. But the demo source for handleBuild implies that there is a way to handle a scene of meshes and mark each one with area type:
// Find triangles which are walkable based on their slope and rasterize them.
// If your input data is multiple meshes, you can transform them here, calculate
// the area type for each of the meshes and rasterize them.
memset(m_triareas, 0, ntris*sizeof(unsigned char));
rcMarkWalkableTriangles(m_ctx, m_cfg.walkableSlopeAngle, verts, nverts, tris, ntris, m_triareas);
if (!rcRasterizeTriangles(m_ctx, verts, nverts, tris, m_triareas, ntris, *m_solid, m_cfg.walkableClimb))
{
m_ctx->log(RC_LOG_ERROR, "buildNavigation: Could not rasterize triangles.");
return false;
}
But i cant see how you would do that… They don’t give any notion as to how one iterate the meshes and assign an area type. Do have any idea how they mean for use to do that at that point in the build process ?
I also see the Convex Volume Tool … Which seems to allow you to set the area type within a extra shape volume… like a box (I don’t see different shape types). Thats a bit different than marking your meshes in Unity Editor and assigning Walkable or Non Walkable or even custom areas and its cost. But it would work pretty good for exporting content from unity where you use the new NavMeshComponents and the NavMeshModifierVolume component which does basically the same thing as the RecastDemo ConvexVolumeTool . So i will definitely look into that.
But it would be really kool if i could somehow mark the original input meshes with area types. Imagine a scene where i have geometry representing a lake or something and i want the exact layout of the lake to be marked with the WATER area and have a higher cost like 10 something
If you just constantly click around, especially longer paths, when the cube is following the path, you can see a slight jitter of the cube as it moves along the path… Any way to smooth that out ???
Hey @Cedric … The previous version of recastjs.cpp … (That i actually use in native code from the Unity editor to generate the navmesh at design time). I created a saveNavMeshData(const char path)* that works great:
// Mackey Kinard
int NavMesh::saveNavmeshData(const char* path)
{
FILE* fp = fopen(path, "wb");
if (!fp) return -1;
// Setup header.
NavMeshSetHeader header;
header.magic = NAVMESHSET_MAGIC;
header.version = NAVMESHSET_VERSION;
header.numTiles = -1;
if (m_navMesh)
{
// Get Navmesh
const dtNavMesh* mesh = m_navMesh;
// Store header.
header.numTiles = 0;
for (int i = 0; i < mesh->getMaxTiles(); ++i)
{
const dtMeshTile* tile = mesh->getTile(i);
if (!tile || !tile->header || !tile->dataSize) continue;
header.numTiles++;
}
memcpy(&header.params, mesh->getParams(), sizeof(dtNavMeshParams));
fwrite(&header, sizeof(NavMeshSetHeader), 1, fp);
// Store tiles.
for (int i = 0; i < mesh->getMaxTiles(); ++i)
{
const dtMeshTile* tile = mesh->getTile(i);
if (!tile || !tile->header || !tile->dataSize) continue;
NavMeshTileHeader tileHeader;
tileHeader.tileRef = mesh->getTileRef(tile);
tileHeader.dataSize = tile->dataSize;
fwrite(&tileHeader, sizeof(tileHeader), 1, fp);
fwrite(tile->data, tile->dataSize, 1, fp);
}
}
fclose(fp);
return header.numTiles;
}
But now that dynamic object require tiles, i need to tweak to support tiles… so i can generate the nav mesh bin file at design time in the editor. Can you please help me fill in the blanks here so save the tiles nav mesh data to disk file so i can reload later using babylon js at runtime.
Her is the skeleton code i created to handle both single mesh and tiles
int NavMesh::saveNavmeshData(const char* path)
{
FILE* fp = fopen(path, "wb");
if (!fp) return -1;
// Setup header.
NavMeshSetHeader header;
header.magic = NAVMESHSET_MAGIC;
header.version = NAVMESHSET_VERSION;
header.numTiles = -1;
if (m_navMesh)
{
// Get Navmesh
const dtNavMesh* mesh = m_navMesh;
if (m_tileCache)
{
// TODO: Save Tiled Nav Mesh
}
else
{
// Store header.
header.numTiles = 0;
for (int i = 0; i < mesh->getMaxTiles(); ++i)
{
const dtMeshTile* tile = mesh->getTile(i);
if (!tile || !tile->header || !tile->dataSize) continue;
header.numTiles++;
}
memcpy(&header.params, mesh->getParams(), sizeof(dtNavMeshParams));
fwrite(&header, sizeof(NavMeshSetHeader), 1, fp);
// Store tiles.
for (int i = 0; i < mesh->getMaxTiles(); ++i)
{
const dtMeshTile* tile = mesh->getTile(i);
if (!tile || !tile->header || !tile->dataSize) continue;
NavMeshTileHeader tileHeader;
tileHeader.tileRef = mesh->getTileRef(tile);
tileHeader.dataSize = tile->dataSize;
fwrite(&tileHeader, sizeof(tileHeader), 1, fp);
fwrite(tile->data, tile->dataSize, 1, fp);
}
}
}
fclose(fp);
return header.numTiles;
}
Shit… didnt even look for a cmake file… I created Xcode and Visual Studio projects so i could create RecastTools.dll and RecastTools.bundle that is a pure native project i use to build the nav mesh at desgin time. So i was not using cmake
i noticed the single sheet nav mesh save code is a little different… it seems you are now adding some sort of recast header BEFORE you write the NavMeshSetHeader.
Hi @MackeyK24
yes, I wanted to preserve backward compatibility with previously saved data.
There is an header named RecastHeader that’s common for tiled or simple navmesh.
It contains the magic number, the version and the number of tiles.
Then, there is the TileCacheSetHeader or the TileCacheTileHeader depending on the magic.
The magic number is NAVMESHSET_MAGIC for simple navmesh or TILECACHESET_MAGIC for tiled ones.
In RecastJS, there is no direct file access, data is read in a buffer (or written) and then processed. That’s what the snippet you posted does.
it copies the data to a buffer and resizes it on the fly. that buffer is returned to JS when everything is in.
For offmesh links, I don’t know. I guess that if it works with simple navmesh, it should work with tiled.
Hello! I need to synchronize the location of agents across clients from the server via socket.io so that the location on all clients is identical at least once every 5-10-30 seconds (not so important), but I cannot force the agents to change their location. Maybe there is a built-in method in RecastJSPlugin, or maybe everything is very simple, but I don’t understand.
I attached an example from the sandbox (the code is of course very poor, but I’m a beginner).
The essence of the example, I created NavMesh, and after some time I try to correct
_navigateP.agents.forEach(m => {
m.mesh.position = new BABYLON.Vector3(…
but it doesn’t lead to anything.
I saw that you can use crowd.agentTeleport(idx, vector), but when teleporting, the movement breaks to the destination
Help me please. https://playground.babylonjs.com/#F9W06U#20 @Cedric, I pray to you)
Yes, agentTeleport will teleport the agent anywhere on the navmesh. It’s also right it will break its navigation. But you can send its destination along its position when synchronizing. Then, in the code, call agentTeleport and agentGoto.
It’s a pity, but on the other hand, thanks to your answer, I discovered a moment that I didn’t foresee I will save the end point in all meshes, and send everyone to their destinations when updating Thank you very much, as always helped
Hi all. I come back from time to time to refine the features of my mini-project
Now I have a need for another refinement of the logic of moving units in my place, and again I need the help of professionals. I dug up my old playground, experimented, but hit a dead end (my code hasn’t improved much in the meantime ).
The problem is this - I added the ability to chase a unit with another unit. The unit continues to move if there is not enough distance to the target (attack distance). Lines 227-237
But I can’t figure out how to make the agent STOP initially when there is enough distance, that is, so that he does not initially go point-blank, where the other unit, but stops. And he continued to move only if the detachment moved away from him.
In simple terms, like in Warcraft3 or Starcraft, chase mode if you click on an enemy unit
help me please
