Mass(Entity) RoadMap/Vision and more questions

Development Platform & Builds
, ,
1

Hi Epic team,

First off, thank you for all the work you’ve put into the Mass (Entity) system. It’s been exciting exploring its capabilities in large-scale simulations and AI systems. We’ve been integrating Mass into our UE5.5 project and have a number of questions - both from our experience and common concerns seen across the community.

1.Roadmap & Vision

  • What’s the current roadmap for the Mass system as it evolves in future versions of Unreal Engine?
  • Any insight into long-term support or target features would help teams plan for deeper integration.
  • Several mass plugins are still marked as “experimental” - from our understanding those were already used in FN experiences in which case they should be in beta status, but if not - when do you estimate they will be promoted to beta status?

2.Entity Lifecycle & Reintegration

  • While removing Entities from Mass is straightforward, what’s the intended path for re-integrating them into a Mass Spawner (or Mass system in general) without fully respawning them?
    • This is important for AI that might leave the system temporarily for gameplay/scripted control and need to return afterward.
  • Are there plans to support more dynamic runtime modifications of entity composition—such as adding/removing fragments on the fly?
    • This would make behavior switching and reactive AI much easier to author.

3.Designer-Facing Tools & Instance-Level Behavior

  • Are there plans to expose more Mass interactions to designers through Blueprints or in-editor tooling?
    • Currently, the system is very programmer-centric with the majority of functions stuck behind C++, with limited per-instance controls.
  • Is Epic considering support for more individualized state machines or state control per entity while keeping performance intact?
    • Designers are often constrained to broader behavioral systems.
  • Will there be better editor-based or visual authoring tools for Mass (fragment viewers, debug overlays, simulation playback, etc.)?
  • Any chance of native Blueprint support for interacting with fragments or triggering behaviors via processors?
    • A Blueprint-first workflow would help hybrid teams.

4.Traffic System & Movement Features

  • Besides lane-based systems, are there plans to support alternate movement types?
    • e.g., ladders, trains, elevators, or path systems for multi-modal traversal
  • Is single-entity goal-driven movement something you’re considering?
    • e.g., sending a Mass agent from A to B and integrating them into traffic along the way
    • This would massively help more dynamic or emergent behaviors.
  • Are there improvements planned for the Traffic Intersection painting tools?
    • The current workflow is a bit clunky, especially with complex intersections.
  • Avoidance behavior in traffic seems inconsistent, and rule configuration via traits/fragments can be opaque. Are there plans to improve reliability or introduce better debugging tools for this?

5.AI Identification, Spawning & Filtering

  • Is there any upcoming support for more robust entity identification/filtering in Mass Spawners—such as performant tagging or classification for spawning and behavior routing?
  • When dealing with duplicate or similarly configured Mass Entity Configuration assets, how can developers ensure correct spawning behavior with MES?
    • This is especially important for multi-agent scenes with variation.

6.Simulation Control, Debugging, & Testing

  • Are there any features planned for deterministic or semi-deterministic simulation playback/debugging for Mass systems?
  • This would help QA teams and designers iterate and reproduce issues quickly.
  • Will Epic expose more Mass-specific profiling or benchmarking tools to help evaluate fragment and processor performance at scale?

7.Replication & Networked Mass Support

  • What is Epic’s vision for Mass in multiplayer contexts—particularly for high-density replicated simulations?
  • Any best practices or plans for built-in support would be extremely helpful.

8.Integration With Other UE Systems

  • How is Mass intended to coexist or interoperate with systems like Physics, Behavior Trees, GAS, or Control Rig?
  • Some systems feel parallel or overlapping, and it would help to know the long-term integration strategy.
2

Steps to Reproduce
steps to reproduce: not relevant

3

1.Roadmap & Vision

What’s the current roadmap for the Mass system as it evolves in future versions of Unreal Engine?

We are working on Mass currently to improve the Observer system to be more performant and robust. There are also plans to add the ability to specify relationships between entities such as parent/child or sibling for the purpose of processor execution. We are also introducing a simplified API for setting up Mass Processors that requires less boilerplate code and improves type-safety while enabling static analysis. 5.6 will also see the MassInsights plugin which adds lots of features for better debugging of Mass and its entities. Other items that we are investigating are serialization, replication, improved parallelism, timeslicing/bathcing processors, adding payloads to MassSignals, entity subqueries, entity disabling, shared fragment improvements, virtual/sparse fragments, and background/long-running processors.

Several mass plugins are still marked as “experimental” - from our understanding those were already used in FN experiences in which case they should be in beta status, but if not - when do you estimate they will be promoted to beta status?

We do use InstancedActors (and thus MassGameplay as it is a dependency) in Lego FN. The InstancedActors plugin was made for Lego FN originally and has been ported out to the engine. I do not know that we have any ETA for when plugins may become production-ready or beta. Some plugins may not ever become fully supported.

2.Entity Lifecycle & Reintegration

While removing Entities from Mass is straightforward, what’s the intended path for re-integrating them into a Mass Spawner (or Mass system in general) without fully respawning them?

Our approach has been to keep the entity while removing the Actor representation when switching LOD. There are means to keep the Actor around, and there is the beginnings of a pooling system for MassActors although it was not used in CitySample. I do not believe we have done much with re-using entities once removed.

Are there plans to support more dynamic runtime modifications of entity composition—such as adding/removing fragments on the fly?

Currently, this is done by using MassCommands to add/remove fragments or tags at runtime. Do you have something else in mind? Are there cases where the commands do not cover what you are trying to accomplish?

3.Designer-Facing Tools & Instance-Level Behavior

Are there plans to expose more Mass interactions to designers through Blueprints or in-editor tooling?

There are not plans for BP currently as it would negate all of the performance gains of Mass. We have discussed some ideas of a BP-lite prototyping for Mass, but there are no plans currently to work on this further. Can you elaborate on what you mean by per-instance controls? Is this changing values on a single entity’s fragments or composition?

Is Epic considering support for more individualized state machines or state control per entity while keeping performance intact?

MassStateTree and StateTree have seen several improvements for 5.6 as we have worked on it for other internal projects. It increases the performance of StateTree and moves more of it off of the game thread. Mass entities using StateTree have their own instance of the StateTree data but share the same fragment containing the StateTree asset. What are you trying to achieve with regards to behavior? It may be something we have not encountered as a use case to consider.

Will there be better editor-based or visual authoring tools for Mass (fragment viewers, debug overlays, simulation playback, etc.)?

There is a new MassInsights plugin coming in 5.6 that also includes updates for the Mass Debugger which will allow to see the fragments and values of any entity in the world and even see the values update in realtime with integration to both Gameplay Debugger and Unreal Insights. You can also see processor dependencies, what processors use a given fragment as well as the mode used, the history of an entity’s composition, and more.

4.Traffic System & Movement Features

Besides lane-based systems, are there plans to support alternate movement types?

We plan to improve MassMovement and MassAvoidance. Some of this may come down to how pathfinding is done and supplying animation to entities. There is a very experimental/debug version of navmesh support in the engine as well as other projects that have attempted to add navmesh support to Mass entities. This should allow for pathfinding through other modes or nav links as would be expected with Actors. For navmesh on moving platforms (boats, trains, floating platforms, etc.), it is something we want to address with navigation in general, but I do not have an ETA on when that work may be done.

Is single-entity goal-driven movement something you’re considering?

We have been working on MassMovements improvements for navigation and avoidance. We do not have any plans to add MassTraffic integration into the movement/path following code.

Are there improvements planned for the Traffic Intersection painting tools?

The current workflow is a bit clunky, especially with complex intersections.

ZoneGraph is currently only in maintenance mode for critical bug fixes with no plans for further development at this time.

Avoidance behavior in traffic seems inconsistent, and rule configuration via traits/fragments can be opaque. Are there plans to improve reliability or introduce better debugging tools for this?

MassTraffic was written specifically for The Matrix Awakens/City Sample projects. I do not believe we will be doing any further work on the plugin as it is not an official engine plugin. We have been working on MassAvoidance and MassMovement improvements, but I do not think remember if the traffic plugin uses MassAvoidance or solely relies on the traffic managers.

5.AI Identification, Spawning & Filtering

Is there any upcoming support for more robust entity identification/filtering in Mass Spawners—such as performant tagging or classification for spawning and behavior routing?

When dealing with duplicate or similarly configured Mass Entity Configuration assets, how can developers ensure correct spawning behavior with MES?

You can set the proportions for each entity config to be spawned and the total entities to spawn. Are you seeing cases where the number or proportion of agents being spawned is not correct? Can you share out any steps to reproduce this?

6.Simulation Control, Debugging, & Testing

Are there any features planned for deterministic or semi-deterministic simulation playback/debugging for Mass systems?

Not at this time. The only deterministic part to Mass currently is the processor execution order. There have been other requests for this, but other items have had a higher priority.

Will Epic expose more Mass-specific profiling or benchmarking tools to help evaluate fragment and processor performance at scale?

Yes. MassInsights should help dramatically with this. It will be included in 5.6. It shows the amount of time spent in various processors as well as what processors are using a given fragment.

7.Replication & Networked Mass Support

What is Epic’s vision for Mass in multiplayer contexts—particularly for high-density replicated simulations?

Serialization and replication are things we plan to address in future development. We do not have any current best practices for Mass replication. We had been waiting for Iris (new networking tech) to be finished, and will need to work with the Networking team towards making MassReplication ready.

8.Integration With Other UE Systems

How is Mass intended to coexist or interoperate with systems like Physics, Behavior Trees, GAS, or Control Rig?

These are areas we are looking into ourselves. Some things will likely not be ported to work with Mass such as BehaviorTree or GAS. We have been working on some internal projects to integrate Mass with other things such as AnimNext, Mover 2.0, and other engine features.

I believe I added information for each question. If I missed one, please let me know as it was unintentional. I am happy to provide some other details as desired.

-James

5 Likes