Behaviors¶
Behaviors are modular steering passes that run each simulation step. Enable them in a SwarmBehaviorProfile assigned to your archetype. Most profiles only need two or three active behaviors.
Finding a behavior setting fast
The Behavior Profile inspector has a search-and-filter bar at the top. See Finding a Setting to search by word or symptom, or to show only the fields you have changed.
Where to start¶
| Game type | Recommended behaviors |
|---|---|
| Default horde | Approach & Press, Personal Space, Idle Motion |
| Survivor-like crowds | Approach & Press (strong), Personal Space (moderate) |
| Melee packs | Approach & Press + Personal Space |
| Flankers / interceptors | Add Predictive Catch-Up |
| Harassers / strafing enemies | Orbit Style + light approach pressure |
| Loose crowd cohesion | Small Hold Nearby Swarm (tune carefully against Personal Space) |
Simple mode field finder¶
The inspector's Simple mode surfaces a condensed set of fields organized into four blocks. Each field links to its full documentation in the Advanced subgroup pages.
| Simple block | Field | Full documentation |
|---|---|---|
| Approach Target | Approach Weight | Approach & Press |
| Approach Target | Combat Standoff | Approach & Press |
| Crowd Spacing | Spacing Strength | Personal Space |
| Crowd Spacing | Spacing Radius | Personal Space |
| Out-of-Sight Fallback | Use Out-of-Sight Fallback | Out-of-Sight Fallback |
Runtime blending¶
Behavior steering is additive — the manager combines all enabled contributions, clamps the result to unit length, smooths it through the movement profile, and multiplies by Max Speed.
Key points:
- Disable, don't zero. Several behaviors cost query budget even when their weight is
0. Disable the behavior fully when you want it gone. - Weights are relative, and they saturate. A weight of
2on Approach against1on Personal Space makes Approach pull twice as hard before the clamp. The combined steering is capped at unit length, though, so once the total reaches that cap raising one behavior's weight only turns the result toward it and pulls the others down — past the cap a weight stops being independent. Tune in small steps and watch the whole mix, not one knob. Agent Spacing ScaleinSwarmSettings—0disables Personal Space entirely. Values above1expand effective spacing radii for approach and orbit.- Approach & Press uses a built-in LOD. When
Use Out-of-Sight Fallbackis on, agents outsideSight Distancefall back to lost-target leader following; with the toggle off (the default), every agent always pursues directly. Reduced and Cheap tier agents reuse cached steering, refreshed every N fixed steps (Approach And Press Reduced Interval/Cheap IntervalinSwarmSettings).
Importance LOD and behavior cost¶
The LOD system (configured in SwarmSettings → Importance LOD) reduces behavior frequency for distant agents:
| Tier | Behavior cost |
|---|---|
| Full | All behaviors run every step |
| Reduced | Personal Space runs at reduced rate and with fewer samples. Approach & Press steering recomputed every N steps (cached between) |
| Cheap | Personal Space even less frequent, further-reduced sample cap. Approach & Press steering recomputed even less often (cached between) |
Tier thresholds are set by Reduced Quality Start Distance (Full → Reduced) and Cheap Quality Start Distance (Reduced → Cheap). Full Quality Agent Cap prevents cost spikes when many agents converge on one target.
Diagnosing behavior problems¶
When a behavior change has no visible effect, or an agent moves in an unexpected direction, use Swarm Debug (Tools > Massive Swarm System > Swarm Debug). It shows the live steering contribution from each behavior for one selected agent, and lets you isolate or mute individual behaviors without editing profile assets.
See Swarm Debug for the full reference.