Node Canvas AI

The Escapists 2 uses NodeCanvas by ParadoxNotion for all AI behaviour. Every NPC's decision-making is driven by a behaviour tree with per-character-type custom tasks and decorators.

Overview

Every AICharacter (and its subclasses: AICharacter_Guard, AICharacter_Inmate, AICharacter_Medic, AICharacter_Dog, AICharacter_MaintenanceMan, AICharacter_JobOfficer, AICharacter_CrowdNPC, AICharacter_Ghost, AICharacter_Generic, etc.) holds a reference to a NodeCanvas Blackboard (m_AIBlackboard) acquired in Awake() via GetComponent<Blackboard>().

The AI runs via ControlledUpdate() (from IControlledUpdate), which is ticked by the UpdateManager each frame. The behaviour tree itself is evaluated separately by NodeCanvas's BehaviourTreeOwner.

Combat Behaviour Pattern

The most notable integration point is the CombatBehaviour blackboard variable (AICharacter.cs:1290–1298):

private const string COMBAT_BEHAVIOUR = "CombatBehaviour";

public void UpdateCombatBehaviour()
{
    AIConfig aiConfig = ConfigManager.GetInstance().aiConfig;
    BehaviourTree combatBehaviour = aiConfig.GetCombatBehaviour(m_CharacterPersonality);
    if (combatBehaviour != null)
        m_AIBlackboard.SetValue("CombatBehaviour", combatBehaviour);
}

Each AI has a Personality.PersonalityType (random 0–3 range). The AIConfig maps personality types to different BehaviourTree assets. This means inmates/guards use different combat trees depending on their assigned personality.

Combat trees are stored as Blackboard variables of type BehaviourTree, and are likely executed by a SwitchBehaviourTree action or a SubTree node that reads the variable.

Personality Types

Personality is assigned at spawn in StartInit():

m_CharacterPersonality = (Personality.PersonalityType)Random.Range(0, 4);

Persisted in save data via CharacterSaveData.m_Personality (int) and restored in DeserialiseCharacter() via UpdateCombatBehaviour().

NodeCanvas Namespaces in the Codebase

NodeCanvas (root)

• Status — Enum: Failure, Success, Running, Resting, Error. Returned by all tasks/nodes to signal execution state.
• ActionListPlayer — Component that wraps an ActionList for manual sequential execution. Supports serialization via JSONSerializer.

NodeCanvas.Framework

• Graph — Abstract base class for all graph types (BehaviourTree, FSM, DialogueTree).
• Blackboard — The data store for AI state. Variables set/get by name. Each AICharacter has one.
• Task — Abstract base for all action/condition tasks.
• ActionTask / ConditionTask — Base classes for game-specific action/condition implementations.
• Node — Graph node base. Has connections, position, and serialized task.
• Connection — Links between nodes.
• BBParameter<T> — A parameter that can be bound to a Blackboard variable.
• Variable — Named, typed value stored in a Blackboard.
• GraphOwner — Abstract component that runs a graph. Concrete: BehaviourTreeOwner, FSMOwner, DialogueTreeController.
• GlobalBlackboard — Shared blackboard accessible across all graphs.
• SyncBlackboard — Synchronizes variables between two blackboards.
• ActionList / ConditionList — Sequential/parallel lists of tasks.

NodeCanvas.BehaviourTrees

• BehaviourTree / BehaviourTreeOwner — The BT graph and its runner component.
• Composites: Selector, Sequencer, PrioritySelector, ProbabilitySelector, FlipSelector, BinarySelector, Parallel, StepIterator, IntIterator, InteractiveObjectIterator
• Decorators: BTDecorator, Filter, Inverter, Repeater, Timeout, Optional, Interruptor, Guard, ConditionalEvaluator, Remapper, Setter, WaitUntil, StateDecorator, RoutineSwitch, MealTimeDecorator, SpeechDecorator, IconDecorator, CombatDecorator, PlayerDisconnectDecorator, TutorialDecorator, RootSwitcher, NodeToggler
• Nodes: ActionNode, ConditionNode, SubTree, NestedFSM
• BTConnection — Custom connection type for the BT.

Game-specific decorators like RoutineSwitch, MealTimeDecorator, CombatDecorator, StateDecorator, SpeechDecorator, IconDecorator show deep integration — the BT nodes are aware of prison routines, combat state, meal times, and character speech/icons.

NodeCanvas.StateMachines

• FSM / FSMOwner — Finite state machine graph and runner.
• States: ActionState, SuperActionState, ConcurrentState, AnyState, NestedBTState, NestedFSMState
• FSMConnection — Transition between states with conditions.

Used for high-level AI state machines (e.g. patrolling → chasing → fighting). The SwitchFSM action and NestedBTState show cross-graph nesting.

NodeCanvas.DialogueTrees

• DialogueTree / DialogueTreeController — Dialogue graph and runner.
• Statement — A speech line with text, audio, and meta. Implements IStatement.
• Nodes: StatementNode, MultipleChoiceNode, ConditionNode, ActionNode, FinishNode, GoToNode, MultipleConditionNode, SubDialogueTree
• DialogueActor / ProxyDialogueActor — Actor references for dialogue.
• IDialogueActor — Interface for dialogue participants.
• SubtitlesRequestInfo / MultipleChoiceRequestInfo — Event structs for UI handling.

Dialogue trees are likely used for quest conversations and tutorial interactions.

NodeCanvas.Tasks.Actions (built-in tasks)

A large library of built-in action tasks covering: movement (MoveTo, Patrol, Flee, Wander, RotateTowards), Mecanim (SetBool, SetFloat, SetInt, PlayAnimation), gameplay (Instantiate, Destroy, SendMessage, SendEvent), blackboard (SetVariable, GetProperty, SetProperty), UI (CameraFader, FadeIn/FadeOut), dialogue (StartDialogueTree, SayRandom), and SwitchBehaviourTree/SwitchFSM.

NodeCanvas.Tasks.Conditions (built-in conditions)

(Condition tasks found in the decomp — used for BT condition nodes.)

Integration Summary