Create AI Controller

[Tapawingo]

Purpose

Create a modular AI Controller that owns decision-making for non-player forces. It should integrate ScenarioData.tasks (TaskDefend/Move/Patrol/SetBehaviour/SetCombatMode/Wait) and unit behaviour (careless/safe/aware/combat/stealth) + combatmode (hold_fire/fire_when_fired_upon/open_fire) so AI agents execute authored task chains deterministically.

Scope

• Task ingestion & queues
  • On mission start, build a per-unit TaskQueue from ScenarioData.tasks (using unit_index, prev_index/next_index).
  • One active task per unit; supports pause/resume/cancel/advance.
• Task semantics (minimal, deterministic)
  • TaskMove → path to point; uses behaviour for caution/speed multipliers.
  • TaskDefend → hold within radius, prefer cover, set observation arc.
  • TaskPatrol → cycle or ping-pong; dwell uses TaskWait.
  • TaskSetBehaviour → set unit behaviour immediately (see mapping below).
  • TaskSetCombatMode → set ROE immediately.
  • TaskWait → block until timer elapses or contact seen (if flag set).
• Behaviour → AI/Movement/LOS mapping (placeholder)
  • careless: fast transit, low cover weighting, long regroup timer, high noise.
  • safe: normal speed, moderate cover weighting, standard regroup.
  • aware: slower speed, high cover weighting, quicker reaction.
  • combat: bounding movement (stub), very high cover weighting, shortest reaction.
  • stealth: slowest speed, maximal concealment bias, limited fire (pairs well with hold_fire).
• CombatMode (ROE) → CombatAdapter
  • hold_fire: never initiate; only engage if explicit fire order (not in this issue).
  • fire_when_fired_upon: return fire only (threat in LOS and hostile shot observed).
  • open_fire: engage viable targets per profile.
• Overrides & precedence
  • ScenarioUnit’s initial behaviour/ROE applied at spawn.
  • TaskSetBehaviour / TaskSetCombatMode override current state.
• AI Controller integration
  • AIAgent consumes TaskQueue → emits intents to Movement/Combat adapters.

Deliverables

• ai/tasks/ScenarioTaskRunner.gd - per-unit runner; processes the chain and raises events.
• ai/AIController.gd (update) - register TaskRunners and tick them in fixed-step.
• ai/AIAgent.gd (update) - apply behaviour/ROE, expose setters, and translate tasks to intents.
• adapters/MovementAdapter.gd (update) - accept behaviour-derived speed/caution multipliers & cover bias.
• adapters/CombatAdapter.gd (update) - enforce ROE.
• adapters/LOSAdapter.gd (update) - support until_contact wait checks.
• Signals (documented in GDScript)
  • task_started(unit_id, task_type: StringName)
  • task_completed(unit_id, task_type: StringName)
  • task_failed(unit_id, task_type: StringName, reason: StringName)
  • behaviour_changed(unit_id, behaviour: int)
  • combat_mode_changed(unit_id, mode: int)
• Acceptance
  • Units with chained tasks execute in order: Move → Defend → Wait → Patrol, etc.
  • Behaviour changes affect path choice/speed; ROE changes affect when units fire (hold/return/open).
  • TaskWait respects duration and until_contact via LOS; TaskPatrol loops correctly.