Home / Projects / Hitz AI Suite

Independent project, published on Fab

Hitz AI Suite

An AI toolkit for Unity 6: behavior trees, state machines, perception, EQS, navigation, crowd avoidance and smart objects on one scheduler, with a visual debugger.

Engine
Unity 6 (6000.3+)
Language
C#
Performance
Burst jobs for EQS
Type
Solo project
Distribution
Fab
The Warden boss mid-slam in the Hitz AI Suite boss fight sample

Overview

Hitz AI Suite gives a Unity studio the systems it needs to build enemies, companions and crowds, designed to work together and to stay fast with hundreds of agents. If you know Unreal Engine's AI framework, the concepts will be familiar.

My role

I designed and built everything: the runtime systems, the editor tooling, the samples and the documentation.

Main responsibilities

  • AI World: one scheduler for every system, with level of detail and a per-frame time budget.
  • Behavior Trees and State Machines for decisions; Blackboards for agent memory.
  • Perception, Navigation, EQS, Crowd avoidance (ORCA) and Smart Objects.
  • A visual debugger and a GraphView behavior tree editor.
  • A bridge to Unity's own Behavior package.

Technical challenges

Run many agents inside a fixed frame budget, let conditions interrupt lower-priority behaviour at once, make agent decisions visible while debugging, and keep unused systems out of the build.

Implementation

The Boss Fight sample shows every system in one encounter. The Warden sees you in a wide cone, but pillars block its view, and it hears your sword. Its behavior tree picks every move by priority: roar at phase changes, recover from staggers, heal when low, otherwise fight. EQS picks where it leaps, where it summons minions and where it searches when it loses you. Minions run a state machine and swarm with crowd avoidance. In the last phase it claims a healing shrine, and hitting it interrupts the heal.

Hitz AI Suite system architecture Perception writes to the agent blackboard. A behavior tree or state machine reads the blackboard and drives EQS, navigation, crowd avoidance and smart objects. Everything runs inside one AI World with a scheduler, level of detail and a per-frame budget. AI WORLD: ONE SCHEDULER, LOD, PER-FRAME TIME BUDGET Perceptionsight, hearing, damage Blackboardper-agent memory Behavior Tree / State Machinepriorities, interrupts EQSpositions, Burst jobs NavigationNavMesh, budgeted CrowdORCA avoidance Smart Objectsclaimed activities
Senses and decisions share one blackboard. The decision layer asks EQS where to go, moves through Navigation and Crowd, and claims Smart Objects. Every system can be left out of a project.
The behavior tree editor showing the Warden combat branch
The Warden's combat branch in the behavior tree editor.

Results

500 agents with perception, behavior trees, EQS and smart objects run at 0.67 ms median and 1.0 ms p95 of AI main-thread time per frame, with no managed allocations per frame. Measured with the included benchmark in a release Windows player on a Ryzen 7 3700X.

Links and media

EQS scoring candidate summon positions
EQS picks summon spots.
Minions swarming with crowd avoidance
Crowd avoidance on the minion swarm.
The Warden surrounded by meteor markers in phase 3
Phase 3: meteors and faster attacks.
Hitz AI Suite 500-agent benchmark summary
500 agents, release Windows build.