Ninja Bot
Ninja Bot is an AI framework plugin for Unreal Engine that introduces a structured approach to perception evaluation and awareness. It is designed to sit between Unreal's AI Perception system and higher-level decision logic, providing a stable and reusable foundation for AI behavior.
The framework focuses on exposing interpreted, meaningful data to existing AI decision systems while remaining fully compatible with Unreal Engine and the Ninja plugin ecosystem.
Main Features
Simplifies AI configuration and promotes reusability.
Introduces explicit perception evaluation and awareness modeling.
Aggregates multiple senses into normalized awareness data.
Exposes meaningful inputs to Behavior Trees and State Trees.
Provides additional gameplay features such as Patrol Points.
Provides a dedicated and optional module for Threat Management.
Integrated with the User Interface, via base UMG widgets and ViewModels.
Integrates seamlessly with Ninja Factions.
Architecture Overview
Ninja Bot is organized around a clear separation of responsibilities between sensing, interpretation, decision support, and presentation. Each layer is explicit, data-driven, and independently extensible.
At a high level, the data flow follows this sequence:
World Stimuli: Raw events generated by gameplay systems, such as sight, sound, damage, and team context.
Sensing and Teams: Detects stimuli and exposes perception and team data without interpretation or decision logic.
Ninja Bot: Evaluates sensed data into normalized perception entries and interprets them into continuous awareness.
Decision Systems: Consumes awareness data to drive AI behavior using Behavior Trees or State Trees.
UI Presentation: Observes awareness state and presents AI intent through ViewModels and widgets.
By separating sensing, interpretation, decision-making, and presentation, Ninja Bot enables AI behavior to be authored in a fully data-driven and reusable way. Multiple bots can share the same AI controller while differing only in their data assets, including perception settings, awareness profiles and optionally, faction configuration.
This non-opinionated design supports both Behavior Trees and State Trees equally, allowing teams to choose the decision model that best fits their project while sharing the same perception and awareness backend.