Is it possible to simulate a large open world in UEFN using layered Data Layers + custom streaming logic?

I’m trying to design a large-scale “open world” style experience inside UEFN, but without using the Landscape system or trying to extend the world space beyond its limits.

Instead, my idea is to build everything as stacked layers in the same coordinate space, using:

  • Custom mesh-based environments instead of Landscape
  • Data Layers + Prefabs (Scene Graph) to represent different “world zones”
  • A fully custom gameplay framework in Verse (AI, inventory, weapons, etc.), without relying on UEFN devices or built-in AI systems

Core concept

The world would be split into modular “regions” (zones), but all occupying the same physical coordinates in UEFN.

When a player reaches a boundary:

  • A cutscene or loading screen triggers
  • Player state is preserved
  • They are teleported/repositioned
  • One Data Layer is deactivated and another is activated

Key idea: stacked but isolated worlds

A critical part of the design is that multiple players can exist in different stacked layers at the same time in the exact same coordinates, but:

  • They cannot see each other
  • They cannot interact or collide
  • Their physics, AI, and gameplay systems are completely isolated per layer

So from the player’s perspective, it behaves like completely separate worlds—for example:

  • Player A is in a “Chicago” layer
  • Player B is in a “New York” layer
  • Both are technically at the same UEFN coordinates, just in different active layers

But they have no awareness of each other, making it feel like entirely separate maps.

AI / systems design

I also plan to implement:

  • Fully modular AI using shared logic templates (no UEFN AI system)
  • State-driven, event-based behavior (no ticking systems if possible)
  • Custom inventory / weapons / gameplay systems entirely in Verse

Questions

  • Is this kind of layered “pseudo-streaming / instanced world” approach actually feasible in UEFN?
  • How does memory budgeting actually behave in this case—is it only what’s actively loaded per player, or still globally constrained even with Data Layers?
  • Are there engine-level limitations that would break this approach at scale (especially with stacked isolated layers)?

I’m basically trying to understand how far a modular, instance-like world architecture can be pushed inside UEFN before hitting hard constraints.

Hey! Really nice idea, from my experience, the overlapping data layers can be done, but it will not work like that. What I got from testing it in UEFN data layers are handled at Server Level, meaning is replicated for all players when one data layer is enabled or disabled.

I was also thinking about doing a big open world like that with data layers, but as Abrizzer said, there is no option to load or unload content of the data layer only for a specific player. So it’s not possible to do large-scale true open worlds with many different locations full of unique objects for many players that could be in many different areas at the same time. Solution would be closing different locations to some logic with counting players in specific areas of the world and lock/unlock access to some areas based on that (so if all players are in the west part of the world, then areas in the east would unload and would not be accessible, while the extended areas further in the west would load and be accessible).

Either that or just depend on streaming and do layout of the whole world based on that which would mean a lot of compromises.