A bolt-on middleware system for game engines that handles large-scale agent simulation using a hybrid GOAP + FSM architecture with built-in LOD (Level of Detail) scaling. The middleware accepts world state inputs and outputs actionable commands for agents, enabling complex AI behavior without being tied to any one engine.
- Plan high-level behaviors using GOAP.
- Execute behaviors using FSMs or micro-behaviors.
- Automatically manage LOD tiers for thousands of agents.
- Run the entire simulation in a highly parallel optimized runtime.
- Provide an adapter layer for integration with external game engines.
- GPU compute offloading.
- Full deterministic multiplayer.
- Visual editing tools for GOAP/FSM (later milestone).
- Networking stack or built-in pathfinding.
- Animation graph integration.
- Unity (first).
- Unreal Engine (after MVP).
- Godot / Stride (optional).
- Handle 500–2000 agents at interactive framerate with GOAP-level planning.
- Minimize memory allocations and support SoA layouts.
- Modular plug-in behavior modules.
- Deterministic mode (planned but not MVP).
- Core Simulation Runtime.
- Planning System (GOAP).
- Behavior Execution System (FSM).
- LOD Manager.
- World State Interface.
- Engine Adapter (Unity target).
- Debug Visualization (console + overlay).
- Sample Scenes and Integration Examples.