Stealth Game Prototype

Download
This short demo is my Unity portfolio project. It's a small game in the stealth genre. It covers most of the Unity systems such as navigation, lighting, animation, audio, character control and is mostly focused on AI.

AI Behaviour Trees (SimpleBT)

I chose to build my own BT system because none of the free BT Unity assets are good and for skills demonstration purposes. It was designed with a KISS principle in mind, although no heavy performance optimisation was done, it's highly extendable and contains most of the basic building bricks.

Basic features

• A single Node task class instead of complicated division of Action, Condition, etc classes.
• Basic Composite Nodes: Selector, Sequence, Parallel, Race
• Interrupts support: Selector.Active node which constantly re-evaluates child nodes with low priority
• Subtrees support
• YAML syntax. There's no GUI editor and BTs are defined in YAML syntax

  LookAround:
    Sequence:
    - Robot.EnableLookAt: true
    - RandomSelector:        
      - Animator.PlayState: { state: "Look Around Fast" , layer: "Actions" }
    - Robot.EnableLookAt: false
 
  Transition_ToAggressive:
    Sequence:
    - Vision.HasTarget: { sensor: $Vision Sensor, target: $Player }
    - Sequence:
      - Success:
        - Sequence:
          - Not:
            - Str.In: { $State, ["Searching", "Aggressive"] }
          - Audio.Play: "Intruder"
      - Animator.SetTrigger: { "Interrupt Action" }
      - Str.Set: { $State, "Aggressive" }

• Node parameterization. One can define a custom Node class with parameters covering all basic types: string, int, float, Vector3, Transform

[Name("Audio.Play")]
public class AudioPlay: Node
{
    private StringParameter name;
    private Audio.AudioManager audioManager;

    protected override void OnStart()
    {
        audioManager = currentContext.GameObject.GetComponent<Audio.AudioManager>();
    }

    protected override Status OnUpdate()
    {
        audioManager.PlayAudio(name);
        return Status.Success;
    }
}

• Blackboard support. One can bind Blackboard values to the BT Node parameter

Vision.HasTarget: { sensor: $Vision Sensor, target: $Player }

• BT Window for visualisation of the current state of the tree

Enemy Robot AI implementation

• Behaviour trees as a core (decision making, plan execution)
• FSM inside of Behaviour tree for better handling Robot states and transitions between them: Idle, Alerted, Aggressive, etc. FSM is emulated via Selector.Active node. This kind of architecture was chosen because I was impressed by the Bobby Anguelov GDC AI talk. In short, BTs are not good with the state management and it's better to combine FSM (state management) and Trees (actual plan execution). I decided to not implement a separate FSM component and emulate it with BT.

Other notable features

2D convex hull collider

It's used in complex vision sensor shape. One can edit it with the gui similar to the polygon collider editor.

AI Patrol Route

This component describes a cycled route for an AI agent. One can assign a tree execution (Wait, Look Around, etc) for a particular position of the route.