[R+D] Step not labelled as walkable to go into a building - thoughts?

[Tanaris4]

All-

So my mesh works for the most part leveraging R+D, but I just ran into a problem where it doesn't recognize a building entrance properly.

Now I'm thinking maybe it's b/c of my parameters, but I don't think I should increase the walkable angle more (b/c then it will think I can go up some mountains when I can't)

Anyone have any thoughts as to why this may be occurring + how to correct?

In game:


RecastDemo:


My generation parameters:

#define GRID_SIZE       (533.0f + (1.0f / 3.0f))
#define TileSize        (533.0f + (1.0f / 3.0f))
#define TileVoxelSize   1800
#define MinRegionSize   20
#define MergeRegionSize 40

#define CellSize        TileSize / (float)TileVoxelSize             // 0.296296296296294

#define CellHeight              0.4f
#define WalkableSlopeAngle      50.0f
#define MaxSimplificationError  1.3f
#define DetailSampleDist        3.0f
#define DetailSampleMaxError    1.25f

#define WorldUnitWalkableHeight     1.652778f
#define WorldUnitWalkableClimb      1.0f
#define WorldUnitWalkableRadius     0.2951389f

#define WalkableHeight      (int)(ceilf(WorldUnitWalkableHeight / CellHeight)) + 3
#define WalkableClimb       (int)floorf(WorldUnitWalkableClimb  / CellHeight)
#define WalkableRadius      (int)ceilf(WorldUnitWalkableRadius / CellSize)

#define MaxEdgeLen          (int)(WalkableRadius * 8.0f)
#define BorderSize          (int)(WalkableRadius + 4.0f)
#define TileWidth           (TileVoxelSize / GridDiv) + (BorderSize * 2)

#define GridDiv 4
#define MaxPoly  1048576 / ( GridDiv*GridDiv)

#define m_vertsPerPoly          6.0f

#define m_keepInterResults      false
#define m_monotonePartitioning  false

Thanks in advance!
~ Tanaris

[DarkLinux]

World of Warcraft also has a step up value. I know Ryuk found that value and he could step over buildings. Its not the same as the walkable angle. So I would think that R+D would have a step up value. But I have never used R+D, so I don't know...

[Empted]

Hello. Everything works smooth for me. Your WalkableClimb is wrong i guess. It's rather big (almost half) related to agent height in WoW.

        public static RecastConfig Default
        {
            get
            {
                var ret = new RecastConfig();
                const float tileSize = Constant.TileSize;
                const int tileVoxelSize = 1800;
                ret.CellSize = tileSize / tileVoxelSize;
                ret.CellHeight = 0.3f;
                ret.MinRegionArea = (int)Math.Pow(6, 2);
                ret.MergeRegionArea = (int) Math.Pow(12, 2);
                ret.WalkableSlopeAngle = 50f;
                ret.DetailSampleDistance = 3f;
                ret.DetailSampleMaxError = 1.25f;
                ret.WorldWalkableClimb = 1f;
                ret.WorldWalkableHeight = 2.1f;
                ret.WorldWalkableRadius = 0.6f;
                ret.WalkableClimb = (int) Math.Round(ret.WorldWalkableClimb/ret.CellHeight);
                ret.WalkableHeight = (int) Math.Round(ret.WorldWalkableHeight/ret.CellHeight);
                ret.WalkableRadius = (int) Math.Round(ret.WorldWalkableRadius/ret.CellSize);
                ret.MaxEdgeLength = ret.WalkableRadius*8;
                ret.BorderSize = ret.WalkableRadius + 8;
                ret.TileWidth = tileVoxelSize;
                ret.MaxVertsPerPoly = 6;
                ret.MaxSimplificationError = 1.3f;
                return ret;
            }
        }

[Tanaris4]

Thanks changing my CellHeight to 0.3f like yours fixed it.

Are there any disadvantages to doing this?

[Bananenbrot]

No, there aren't. It's just a higher resolution regarding your walkable height, which is always rounded up for conservatity (?). This means, in your current configuration you have a walkable height of effectively 2.0f (ceil(~1.7f / 0.4f)*0.4f), while with a cell height of 0.3f it can be reduced to 1.8f.
However iirc, ~1.7f is too small and Empted's 2.1f is much more likely to fit the walkable height of a Tauren.
Note that 0.3f is particularly "fitting" for 2.1f.