Best practices for handling AI behavior trees in large open-world maps?

Hello everyone,
I’m developing an open-world game in Unreal Engine and using behavior trees to control AI characters. As the map gets larger and there are more AI agents active, I’m noticing performance drops and some AI behaving unpredictably.

Could anyone share best practices or optimization techniques for managing AI with behavior trees in large-scale environments? Also, how do you efficiently manage AI updates without sacrificing performance?

Thanks for any advice or resources!

That which does not exist consumes the least resources. :grimacing:
Therefore, the only thing that can be done is to disable the behavior tree for remote NPCs.

Depending on the game’s features, you need to come up with a life simulation system for remote NPCs. In the simplest version, when approaching an NPC that has been unloaded, select a random result of an action that the player has not seen.