Generation Settings

What generation mode should volumator use to create the volume? Each mode uses a different method to create volumes and will have a large impact on the output volume and how Volumator interacts with it.

Perimeter

The normal, default volume mode is 'Perimeter' mode. This mode attempts to find the perimeter from the detected geometry and then attempts to order those perimeter points so they're sequential. 

Perimeter mode works really well for most shapes, although it can struggle with hyper concave or 'spikey' shapes as finding the perimeter can be challenging.

The majority of the documentation assumes you're using Perimeter mode. Other modes are called out specifically and below we'll cover the main high level differences. 

Two Pass

This is an evolution of Perimeter mode. It runs perimeter detection twice (amongst other tweaks) resulting in a much lower chance of broken volumes. The downside is that the volume will be slightly less accurate and more simple, it also takes slightly longer to process (although this is pretty negligible). It's a great solution if Perimeter mode is struggling to build valid shapes.

Voxel

Voxel mode uses an entirely different generation method that can be useful in cases where the Perimeter mode is struggling or if you're working with blocky spaces. Each detected location, plus the path that was taken during piloting or droning, is plotted in a grid. We then add a volume 'block' at each of these locations, creating a blocked out impression of the detected space. 

The result is a volume that is made up entirely of 90 degree angles. It therefore can work incredibly well for blocky level design, such as corridors and tunnels. Obviously it works less well for curved spaces.

The size of the voxels (blocks) is determined by the Resolution setting. 

Voxel mode has a number of advantages over perimeter mode: 

• zero warnings and errors - as every detected point gets a large block of volume, coverage is always guaranteed

• volume is more reliable - as the angles are always 90 degrees

• less chance of integrity issues - avoids the above perimeter 'jumps'

There are some downsides though:

• volumes are always slightly bigger than the shape, so overlaps into other areas may be a problem depending on your context

• curved shapes obviously aren't handled well

• the interior section requires the detection actor to have passed through it, so it's important to be thorough with the interior during detection or holes will appear in the interior of the volume. There is a hole filling operation to patch up interior holes, see below. A 'voxel hole-filling' setting can be found in Generation-Advanced, check the documentation page for more info.

• generation can be slow at lower resolutions and because ALL validation attempts are run (see below), a low resolution generation with validation can take a while. Lower validation attempts or increase Resolution to speed things up

The following features are NOT supported or make no difference in voxel mode:

• Expand - because it's not necessary, the volume will already be expanded to fill the detected points

• Simplify - uses angle differences, but voxel mode always uses 90 degree angles so can't be simplified further

• Simplify concavities - not necessary

The validation system also works differently. As the perimeter mode validation uses the waring/error results, we can't use it for voxel mode because there will never be any. So instead, the validation system will rebuild the volume with increasing resolution settings and will pick the volume with the least amount of voxels required. As the detected points line up with the grid of voxels differently at different voxel sizes, it's possible that even though validation increases the resolution setting (voxel size) there's a chance that it fits the locations better, and this is what the validation system does in voxel mode. It works surprisingly well, trying to find the ideal voxel size. 

While there are some caveats to voxel mode, it can work incredibly well for certain, more blocky scenarios, or times where the volume size being bigger than the space isn't a problem.

There is also the special Vertices-Only mode. This is used by separate the Start Building Vertices-Only button (see Build panel page). You cannot select it in the Generation Settings because a Vertices-only volume is locked to that mode and cannot switch to the others, or vice versa. 

The height of the volume that is generated. We update this value when starting pilot or drone modes, by detecting the height at the start area. Effectively this is the minimum volume height when you're building a volume, as if an area is detected as having a higher height, then it'll increase this Volume Height to match what was detected. 

Note the 'Regenerate changes Height' variable in the Advanced Generation settings; it will determine if this variable impacts regenerations. By default it is off, to save you accidentally changing the height of volumes during regeneration. With it enabled, regeneration of a volume WILL change the height of the regenerated volume to this value. 

A key setting, this dictates the definition and complexity of the volume. It is measured in Unreal units, which means increasing the Resolution setting makes a generated volume more simple, less accurate. 

In Perimeter (normal) mode, it determines the minimum spacing between vertices; e.g. 200 Unreal units is the minimum distance between volume edges. 

In Voxel mode, it determines the voxel size, e.g. each voxel block is 200 Unreal units in width and length.

In Vertices-only mode is has no affect. 

As it is in distance, as this value is increased the complexity is reduced. The CPU cost of the volume is reduced as complexity is reduced, but the volume becomes less accurate to the space. High Resolution (low complexity) can be problematic on curved areas where accuracy is important. 

As this setting is so critical to the accuracy, the Edit tab also has a Resolution tweaker, allowing you to quickly manipulate and try out different resolution settings to see how it affects the shape.

In Perimeter mode, under certain conditions, especially with curved and concave shapes, increasing the resolution too high will cause the integrity of the shape to fail and result in poor quality volumes. See below for an example of how complex shapes are affected by resolution changes.

Resolution is such an important setting for volume integrity, accuracy and CPU cost that there are two validation strategies that focus solely on tweaking the Resolution value to perfect it.

While building, Volumator will determine if the area it's detected is way too large for the current settings. If this happens, it'll update the Min Point Distance and Resolution settings to something more appropriate. This is to stop Volumator getting overwhelmed with data or cause your PC to become sluggish. 

As mentioned in the Voxel Mode section, when making voxel volumes Resolution controls the voxel, or block, size.

We can expand, or inflate, the volume by pushing out each face. This isn't a simple 'scale' control, it expands the space of the volume by pushing each wall outwards away from itself. This is done by detecting and storing the normal (angle) of the object the scan hits during the detection phase, in cases where we haven't hit an object, we calculate the normal using adjacent points. 

This means we can adjust the size of the volume even after we've reloaded the level. Expanding becomes important when you want to reduce the complexity of a volume, as you can push the volume further into the walls as it less perfectly reflects the shape of the area. 

Expand is designed to give you a little buffer space to ensure the volume covers the space, it is not designed to massively change the shape. It can start to misbehave at extreme settings, especially on concave or sharp edges. This is because you may be pushing points out so far that they now 'cross over' the location of their neighbours. 

Expand has no affect in Voxel mode as it's not necessary. 

Resolution is our main method to control volume complexity. However, we also have two angle-based simplification processes that run after Resolution. 

'Simplify' removes any perimeter points that are within an angle threshold of adjoining points. This means if we have three perimeter points in a line with almost no angular change between them, the middle point is likely unnecssary and can be culled.

The variable is measured in degrees, so a value of 1 means if three points are in a line with the middle line being less than 1 degree off the line, the middle point will be culled.

Increasing this will reduce complexity and remove redundant data. However, increasing it too far will start reducing points that contribute to the shape of the space and you will find the volume not representing the space well. 

As this is an important control that benefits from adjustment, it's one of the easily tweakable controls in the Edit panel. 

The default setting is set to remove unnecessary points in a line without impacting the shape much. 

Note that the shape of the volume and the Resolution setting will also impact how the Simplify feature reacts, the values are very interdependent. 

While Resolution and Simplify may appear to have similar effects, they work in very different ways and it's worth experimenting to see which suits your use case. Increasing Resolution creates a coarser scan, while increasing Simplify culls points based on their angle.

Simplify has no affect in Voxel mode as all angles are at 90 degrees.

The second simplification process will ONLY cull points that it deems as being concave; meaning it will cull small volume areas that cut inward into the volume. This can be beneficial as small concavities in an area are often just the result of a bumpy/lumpy surface and do not represent significant areas to be represented... they can often then be discarded without worrying about causing problems. 

Effectively, this simplification process will never make the volume SMALLER or eat into the room, it will only ever increase the volume's size. This makes it a safer way to simplify the volume.

However, if you are dealing with highly complex shapes or very thin walls between areas (e.g. glass walls between adjacent rooms) then you may want to minimize this as it could cause the volume to bleed into the next room, but otherwise it's usually fairly safe to use and is set to a small amount by default.

It's a normalised value between 0 and 1, so tweak and see the impact. Even at full value, it is clamped to ensure it never removes a lot of concavity. 

Simplify Concavities has no affect in Voxel mode.

Enabling this will force the generated volume to be a convex shape. It does this by 'filling in' any concave areas, sometimes called 'gift-wrapping', by utilising a convex hull over the shape. This has the benefit of greatly simplifying and optimising the volume, reducing complexity and having generally better integrity and reliablility.

Convex shapes are those that never cut inward. Examples of convex shapes include:

• square

• cylinder

• hexagon

Concave shapes are the oppsite; shapes that do at some point cut inward. Examples of high level shapes would be:

• L shape

• V shape

• pacman or pizza with a slice removed - the removed slice being the concave area

In most cases concavities detected by Volumator are usually smaller and often negligible enough to be discarded

By enabling this setting, you can easily try out the convex volume if you're having issues with a concave shape, e.g. some game systems don't like concave shaped volumes. 

Note that there is a 'Detect Convex' feature explained in the Advanced settings that will automatically make the volume convex if the system detects that it's a convex enough. This means you don't need to worry about setting this manually for each volume. Instead, the setting described here is a brute force option for you for comparison or override purposes.