UEDEV  /  3D Navigation AI  /  Crowds  ·  Updated

v3.6.0 Use case · Mass crowds

Mass Entity flying crowds for Unreal Engine 5

Drone swarms, bird flocks and fish schools in true 3D through Unreal's Mass ECS — 10,000+ agents on per-agent paths or one shared flow field, with crowds that perch, scan, engage, investigate and flee.

Mass Entity agents
10,000+
Mass traits
8
Mass StateTree nodes
13
Flow-field sampling
O(1)
A drone swarm sweeping over a city skyline, driven by a shared 3D flow field
One shared flow field carries the swarm over the skyline
01 · Two modes

Per-agent paths or one shared flow field

Crowds heading the same way sample one shared CPU flow field in constant time, while the agents that matter get their own asynchronous paths. Auto-registered processors handle request, solve and steer, so you never call the solver yourself. A crowd's first flow field is built off the game thread, so a crowd spawning into a streamed world does not cost one long frame.

Close view of a Mass Entity drone swarm flying through a city in Unreal Engine 5
Close fly-through of the swarm
02 · Behaviours

Crowds that behave, not just move

Ten Mass StateTree tasks and three evaluators let crowds perch, scan, engage, investigate and flee. Agents in a crowd share the work: they sweep an area sharing the viewpoints, close round an intruder the moment any of them sees it, and no two take the same perch, cover or vantage point.

A large school of fish turning together above the reef in Unreal Engine 5
A school turning together without overlapping
03 · Density

Dense crowds that do not overlap

Hold Avoidance Between Scans keeps separation on every frame, and Contact Push pushes apart bodies that overlap — including agents that have arrived — so a crowd no longer piles onto its goal. A crowd agent pushed off the flow field's covered space steers back instead of freezing, and schooling works for Mass crowds as well as pawns.

A crowd of drones rising floor by floor through a skyscraper in Unreal Engine 5
A swarm climbing through a tower interior
04 · Setup

One trait, no code

Add a trait to a Mass Entity Config and the crowd navigates in full 3D, with no per-agent controller and no tick function. Eight traits cover movement, guidance, goals, perception, behaviour status and rendering, and an editor validator flags illegal combinations — the one rule is exactly one movement trait per agent. Crowds render through instanced meshes with LOD representation.

Free demo levels

See it in the demo levels

These levels are part of the 57 free demo levels you download from inside the editor (LCM Nav3D menu › Get the Demo Levels).

Hunter drones weave between shipping container stacks while chasing a runner

Hunter Pack

Five Mass drones hunt as a pack through the container terminal.

  • Mass StateTree
  • Engage
A Mass Entity school of fish streams over the coral, parting round the pinnacles

Reef School

A school of forty follows its lead fish on the flow field.

  • Schooling
Sentry drones patrol a depot with the navigation octree and AI Perception sight cones drawn

Sentry Swarm

Six Mass drones guard a depot as one, then rest on the pads.

  • Scan Area
  • Perch
A drone crosses the archive hall of the warehouse demo level

Warehouse Security

Four Mass security drones watch, sweep, engage and search.

  • Investigate
Starfighters exchange laser fire in the Space Dogfight demo for Unreal Engine 5

Raider Swarm

Eighteen raiders sweep the asteroid belt and close round fighters.

  • Engage
A crowd of drones rising floor by floor through a skyscraper in Unreal Engine 5

Vertical Skyscraper

A crowd threading a tower interior, floor by floor.

  • Flow field
FAQ

Common questions

How many agents can navigate in 3D through Mass Entity?

10,000 or more. Crowds can use per-agent asynchronous paths or one shared flow field that the whole swarm samples in constant time, and they render through instanced meshes with LOD.

Do I need C++ to use Mass crowds?

No. Add the traits to a Mass Entity Config; an editor validator flags illegal trait combinations, and Mass StateTree tasks drive the behaviour.

Can crowds use behaviours like perch and engage?

Yes. Ten Mass StateTree tasks and three evaluators let crowds perch, scan, engage, investigate and flee, and agents share viewpoints, perches and cover so no two take the same one.

Read more in the free documentation:

Give your AI the whole sky.

LCM 3D Navigation AI — true 3D pathfinding and volumetric navigation for Unreal Engine 5, built and supported by a working game developer.

Get it on Fab ↗ Explore LCM 3D Navigation AI →