Satisfactory Under the Hood: Unreal, Significance Manager & Abstracted Belts
Engine & Tech Stack
Engine / language: Unreal Engine (shipped on UE4, later upgraded to UE5), with gameplay in C++ on Unreal's Actor/Component framework.
Map representation: one large, hand-crafted 3D open world (not procedural), streamed with Unreal's level streaming and World Partition rather than a tile grid.
Build system: factories, belts and pipes snap onto the world, each backed by an Unreal Actor.
Core AI & Decision Making
Creatures: wildlife uses lightweight Unreal behaviour (roam, flee, attack) on the NavMesh; there is no deep NPC society, the simulation is logistics rather than agents.
Dispatch: trucks and tractors follow player-recorded paths on autopilot, while trains route over the rail-network graph using block signals.
State Management & Data Architecture
Model: Unreal Actors and components; world state is the sum of placed buildings, belts, pipes, the power grid and inventories.
Belt items: items on conveyors are not individual Actors, they use a compact, abstracted representation so long belt chains stay cheap.
Persistence: the entire world (every building, item and vehicle) is serialised into the save.
Tick Scheduling & Performance Tricks
Significance Manager: Unreal's Significance Manager lowers update rates for distant, less-important machines so far-off factory sprawl costs less.
Bulk simulation: conveyor and pipe contents are advanced in bulk rather than per-item-per-frame, and dedicated servers offload simulation from clients.
Pathfinding & Spatial Systems
Creatures: Unreal NavMesh pathfinding.
Trains: graph pathfinding over the rail network with block signals; road vehicles follow recorded splines.
Spatial: World Partition and spatial hashing drive culling and streaming across the large map.
