Tutorial 07 - Authoring true-3D tactical EQS

TL;DR: LCM Nav3D ships EQS generators and tests that work in true 3D - candidate points anywhere in the navigable volume, scored by volumetric SVO line-of-sight, cover, threat-exposure, LCM Nav3D-path-distance and height. Author a query from these nodes exactly like stock EQS; the only differences are the Apex:-prefixed generators/tests and that "distance" means LCM Nav3D path, not Euclidean or navmesh.

The LCM Nav3D EQS node catalog

Generators (UEnvQueryGenerator, produce 3D points):

Tests (UEnvQueryTest, score/filter):

Contexts (UEnvQueryContext):

Step 1 - Author a sniper-perch query

Create an EQS query EQS_FindSniperPerch:

1. Generator: LCM Nav3D | Hemispherical Ring 3D - GenerateAround = the threat context (Known Threats / Threats From AI Perception), RadiusMin/Max = your engagement band, Upper Hemisphere Only = true (perch above the threat).
2. Test: LCM Nav3D | Line Of Sight 3D (Context = threats) - Score, prefer should-have-LOS - a perch must see its target.
3. Test: LCM Nav3D | Cover Score 3D (Context = threats) - Score↑ - partial cover is good.
4. Test: LCM Nav3D | Height Advantage (Context = threats) - Score↑ - favour high ground.
5. Test: LCM Nav3D | Reachability 3D (Context = Querier) - Filter - the agent must be able to get there.

The winning item is a reachable, high, in-cover position that still has line of sight to the threat - a sniper perch, chosen in full 3D.

Figure 07.1 - The anatomy of a Nav3D query. One generator produces candidate points, then tests score or filter them, top to bottom. Here Apex: Points In SVO Volume is emitting unblocked 3D points across an 800×800×400 half-extent at 200 spacing - unblocked being the part a floor-plane generator cannot do, because it never had a third axis to be blocked in.

The two tests are colour-coded by role in the graph: a filter removes items, a score ranks them. The Details panel on the right belongs to the selected test:

• Test Purpose - Score Only, Filter Only, or both. This is the switch that decides whether a failing item is penalised or deleted.
• Context - what the test measures against. EnvQueryContext_LcmKnownThreats here, so line of sight is evaluated toward the threats the agent knows about.
• Agent Radius - how wide the thing that has to fit is. Leave this at your real agent size; a query that scores a perch a 50 cm drone can reach tells you nothing about a 200 cm gunship.
• Scoring Equation and the Preview graph - the curve mapping raw measurement to score. The preview updates live, so you can see a Linear ramp become a Square bias before you ever run the query.

Step 2 - The EQS Testing Pawn

Drop an EQS Testing Pawn in the level, set its QueryTemplate to EQS_FindSniperPerch, and place threats so the contexts resolve. The candidate items render as colour-graded balls; tweak generator/test parameters and watch the scores update live.

UE's stock EQSTestingPawn is all you need - the LCM Nav3D nodes render through it normally.

Step 3 - Wire the result into a Behavior Tree

1. A Run EQS Query task/service runs EQS_FindSniperPerch and writes the best item to a blackboard Vector key.
2. A MoveTo (or the LCM Nav3D FlyTo BT task) moves the agent to that key.
3. Re-run on an interval (service) so the perch updates as threats move.

(StateTree: run the query from a task that stores the location, then transition to your LCM Nav3D move task - see Tutorial 05.)

The five shipped example queries

Recipes you can build now from the catalog above:

Related

• Tutorial 08 - perception-driven threat sources (the auto-populated threat context).
• Tutorial 05 - StateTree integration (the move side).
• ULcmTacticalQueryLibrary - the same primitives as Blueprint nodes (Is Segment LOS Clear, Find Best Cover, Find Apex Path Sync, ...).