LCM 3D Navigation AI- Infinite 3D Pathfinding & Volumetric Navigation for UE5(Flying, Swimming & Crowd AI)

Distribution Product Discussion
1

Next Gen 3D Navigation - Highly Optimized True Volumetric Navigation for Unreal Engine

Overview

Volumetric navigation for Unreal Engine. Sparse voxel octree pathfinding for flying, swimming, and free-floating agents from a single agent to crowds of over ten thousand.

Description

LCM Nav3D is a full-3D navigation system designed for agents that move through space rather than across it. Instead of a flat, walkable navmesh, it carves your level’s empty air into a tree of boxes a Sparse Voxel Octree (SVO) - and paths through the free space. This gives flying, swimming, and free-floating agents genuine vertical freedom through interiors, around structures, and across open terrain.

Engineered to seamlessly integrate across the entire Unreal AI stack, it provides per-agent control at full parity for Behavior Trees and StateTrees, while Mass Entity extends the same solver to massive crowds. Pathfinding executes asynchronously on the CPU, off the game thread, and octree builds are GPU-accelerated with automatic CPU fallback. The solver is completely deterministic, ensuring identical results across runs for replays, lockstep networking, and automated testing. A GPU path solver is in experimental validation as part of our research program and will ship as a free update.

Important Links

Feature Highlight

  • Core Volumetric Navigation: Provides true volumetric 3D navigation by carving empty air into a Sparse Voxel Octree (SVO).

  • World Support: Supports both finite worlds and streaming infinite worlds. Features chunked generation around agents as they move.

  • Solvers & Smoothing: Includes A* (fastest), Theta* (most direct), and Lazy Theta* (balanced default) solvers. Features raw, linear shortcut, and curved spline path smoothing.

  • Mass Entity Crowds: Pathfinds crowds of 10,000 and beyond through the Mass ECS. Offers per-agent asynchronous paths or a shared flow field sampled O(1) by the entire swarm.

  • Behavior Tree & StateTree: Provides fully supported Behavior Tree and StateTree node sets that produce identical flight behaviors.

  • Tactical EQS in True 3D: Ships with true-3D EQS generators and tests that score real volume instead of a floor plane.

  • AI Perception: Includes an AI Perception bridge that wires stock UE perception into volumetric threat tracking and cover evaluation.

  • Dynamic Obstacles: Restamps the octree as actors move so agents can reroute live around moving geometry.

  • 6-DOF Flight & Swim Locomotion: Features multirotor, bird, and fish archetypes with banking, inertia, turn-rate limits, and arrival deceleration.

  • Hybrid Recast Integration: Supports walking agents on Recast and flying agents on the SVO in the same level using a Z-band filter.

  • Production Ready: Includes Smart Objects, Nav Links, volumetric NavArea modifiers, Gameplay Debugger integration, and server-authoritative replication.

Documentation & Samples

  • Delivered as complete C++ source, fully exposed to Blueprint with no C++ required to use it.

  • Includes 18 demonstration maps covering Behavior Tree, StateTree, and Mass Entity scenarios.

  • Features a comprehensive 124-page illustrated PDF manual, 11 step-by-step tutorials, and a full API reference.

Technical Details

  • Code Modules: LCMNav3D (Runtime), LCMNav3DEditor (Editor)

  • Supported Platforms: Windows (Win64), Linux,PS5,Xbox

  • Network Replicated: Yes, server-authoritative

  • Source Included: Yes

  • Required Engine Plugins : StateTree, GameplayStateTree, SmartObjects, MassGameplay, MassAI, and DataValidation.

2

:rocket: Tech Dive Update: 10,000 Flying Agents & Dynamic 3D Pathfinding

Hey everyone,

I’ve just put together a deep-dive technical breakdown showing LCM Nav 3D in action. If you’ve been wondering how the system scales or how it handles complex environments without a traditional NavMesh, this covers it all.

Watch the full breakdown here:

:hammer_and_wrench: Key Technical Highlights in this Video:

  • Massive Scale: Watch 10,000 agents flying and swimming simultaneously using the same underlying solver.

  • Sparse Voxel Octree: A look at how the environment is queried in true 3D space, allowing agents to climb, dive, and thread gaps that a 2D floor plane simply can’t represent.

  • Dynamic Rerouting: See how agents instantly recalculate mid-flight when obstacles are moved in real-time.

  • Seamless Integration: Dropping a ‘Fly To’ task into a Behavior Tree, or utilizing Mass Entity Configs for massive crowds.

Whether you are building underwater columns, dense aerial dogfights, or multi-level spatial puzzles, the entire volume is fully navigable.

I’d love to hear your thoughts! Drop any questions below about how the C++ source or the 61 blueprint nodes work under the hood.