Skip to content

Phase 63: Cross-Module Feedback Loops

Status: Complete Tests: 74 new (6 test files) Source files modified: 6 (zero new source files)

Overview

Wires four cross-module integration gaps — feedback loops where one system's output should drive another system's behavior. Closes the systemic build-then-defer-wiring pattern identified in the Block 7 brainstorm.

Implementation

Step 0: CalibrationSchema Infrastructure

Added 4 new fields to CalibrationSchema: - enable_event_feedback: bool = False — gates medical/maintenance event consumption - enable_missile_routing: bool = False — gates MISSILE type inference for guided launchers - enable_c2_friction: bool = False — gates comms loss → order freeze - degraded_equipment_threshold: float = 0.3 — fraction broken equipment → degraded logging

All default False for zero behavioral change to existing scenarios.

63a: Detection → AI Assessment

FOW sensor/signature wiring — replaced hardcoded empty data in battle.py's FOW update block: - "sensors": [] → ctx.unit_sensors.get(_u.entity_id, []) — real sensor data flows to detection - "signature": None → _get_unit_signature(ctx, _eu) — real signatures from sig_loader - Added _get_unit_signature() helper with graceful fallback to None on any error - Assessment path already reads FOW contacts when enable_fog_of_war=True — now detection actually produces real contacts

Deviation from plan: Plan called for modifying c2/ai/assessment.py and enabling FOW in modern scenarios. Assessment.py already consumed FOW contacts (Phase 53a); the actual gap was upstream — FOW update block passed empty data. No scenario YAML changes needed since all flags default False.

63b: Medical → Strength & Maintenance → Readiness

Event subscription in engine.py — when enable_event_feedback=True: - Subscribes to ReturnToDutyEvent → calls unit.restore_crew_member(member_id) (to MINOR_WOUND, KIA permanent) - Subscribes to EquipmentBreakdownEvent → marks equipment.operational = False - Subscribes to MaintenanceCompletedEvent → restores equipment.operational = True - Equipment breakdown handler checks degraded_equipment_threshold and logs when exceeded

Unit.restore_crew_member() added to entities/base.py: - Restores injury state (default MINOR_WOUND, not HEALTHY) - Returns False for KIA (permanent) or unknown member_id

Deviation from plan: Plan called for formal DEGRADED unit status with Pk/movement/detection penalties. Implemented as threshold-based logging; formal enum value deferred.

63c: Checkpoint State Completeness

Added 4 previously missing engines to both get_state() and set_state() engine lists in scenario.py: - comms_engine - detection_engine - movement_engine - conditions_engine

Deviation from plan: Plan described registering with CheckpointManager. Actual pattern is the engine tuple list in get_state()/set_state() — same mechanism used by all other engines.

63d: MISSILE Routing & Comms → C2

MissileEngine instantiation in scenario.py _create_engines(): - MissileEngine(dmg_engine, bus, combat_rng) — follows same pattern as all combat engines - Added missile_engine field to SimulationContext and result dict

MISSILE type inference in battle.py engagement routing: - When enable_missile_routing=True: weapon category MISSILE_LAUNCHER + guidance != NONE → EngagementType.MISSILE - missile_engine passed to route_engagement() call

MISSILE handler in engagement.py route_engagement(): - Calls missile_engine.launch_missile() with proper args (launcher_id, missile_id, positions, ammo, timestamp) - Returns not-engaged with aborted_reason="no_missile_engine" if engine absent

C2 friction gate in battle.py OODA DECIDE handler: - When enable_c2_friction=True: checks _compute_c2_effectiveness(ctx, unit_id, side) - If effectiveness < c2_min_effectiveness (default 0.3): skips DECIDE phase, unit operates on last orders

Files Modified

File Changes
stochastic_warfare/simulation/calibration.py 4 new fields (3 enable flags + 1 threshold)
stochastic_warfare/simulation/battle.py FOW sensor/signature fix, _get_unit_signature helper, MISSILE type inference, missile_engine in route_engagement call, C2 friction gate
stochastic_warfare/simulation/engine.py _register_event_handlers(), 3 handler methods, unit lookup helper, degraded equipment threshold check
stochastic_warfare/simulation/scenario.py missile_engine field + instantiation, 4 engines added to get_state/set_state
stochastic_warfare/combat/engagement.py MISSILE type handler in route_engagement()
stochastic_warfare/entities/base.py restore_crew_member() method

Test Files

File Tests
tests/unit/test_phase_63_infra.py 13
tests/unit/test_phase_63a_detection_ai.py 12
tests/unit/test_phase_63b_event_feedback.py 14
tests/unit/test_phase_63c_checkpoint.py 10
tests/unit/test_phase_63d_missile_comms.py 13
tests/unit/test_phase_63_structural.py 12

Deferrals

  1. MissileEngine per-tick update — update_missiles_in_flight() advances in-flight missiles. Launch ships; flight-to-impact resolution deferred.
  2. MissileDefenseEngine intercept — Intercepting in-flight missiles requires missile-as-contact detection. Deferred.
  3. Formal DEGRADED unit status — Equipment breakdown logs threshold but no formal enum value or automatic Pk/movement penalties.
  4. FOW confidence-weighted assessment — Assessment uses simple contact count. Sophisticated confidence mapping deferred.
  5. RTD to HEALTHY — RTD restores to MINOR_WOUND. Full rehabilitation deferred.
  6. Logistics event feedback — SupplyDeliveredEvent/RouteInterdictedEvent/ConvoyDestroyedEvent remain unsubscribed.

Postmortem

Scope: Over (74 tests vs planned ~45)

More tests than planned due to additional structural verification and edge case coverage. All 4 planned subsections delivered. Three scope reductions: no assessment.py modification needed (gap was upstream), no formal DEGRADED enum, no scenario YAML changes.

Quality: High

  • All 74 Phase 63 tests pass
  • Full suite 8483 passed, 0 failed
  • Calibration coverage test caught degraded_equipment_threshold as unconsumed — fixed immediately
  • Three test quality issues found and fixed during postmortem (logic error in structural test, empty test body, fragile .index() call)

Integration: Fully wired

  • All 10 integration checks pass (missile_engine end-to-end, all 3 flags consumed, event subscription verified, checkpoint lists complete, helper called, C2 friction wired)
  • No orphaned code, no dead imports
  • Zero new source files — all changes in existing modules

Test quality notes

  • Mix of behavioral (unit tests with real objects) and structural (source string assertions)
  • Structural tests are inherently fragile to refactoring but catch regressions fast (Block 7 pattern)
  • Edge cases well-covered: KIA permanence, unknown IDs, missing engines, empty equipment

Deficits: 6 deferrals (see above)

All are planned future work, not bugs. MissileEngine flight resolution is the most impactful deferral.

Lessons Learned

  • FOW gap was upstream (battle.py passed empty data), not downstream (assessment.py) — always trace the data flow, not just the consumer
  • Calibration coverage test is an excellent safety net — caught unconsumed field before commit
  • MissileEngine constructor signature differs from how COASTAL_DEFENSE/AIR_LAUNCHED_ASHM call it (pre-existing bug in engagement.py lines 412-432: missing launcher_id/missile_id args)