Smart Airport Digital Twin & Operations Management System

A full operational digital twin of an airport, built end-to-end: every flight, passenger, bag, gate, runway, counter, security checkpoint, ground vehicle, and staff resource lives in a shared SQLite database and evolves tick-by-tick through realistic, interconnected state machines — per the "Full Integration Requirement" (bad weather -> delays -> gate conflicts -> boarding delays -> runway congestion -> reduced throughput).

What's implemented

• Flight lifecycle: SCHEDULED -> CHECKIN_OPEN -> BOARDING_PREP -> BOARDING -> READY_FOR_PUSHBACK -> PUSHBACK -> TAXI_OUT -> RUNWAY_QUEUE -> TAKEOFF -> CLIMB -> CRUISE -> DESCENT -> LANDING -> TAXI_IN -> AT_GATE_ARRIVED -> DEBOARDING -> CLEANING -> TURNAROUND -> COMPLETED.
• Passenger journey: booking -> check-in (manual / kiosk / mobile / premium) -> baggage drop -> security -> immigration (international only) -> gate area -> boarding -> in-flight -> arrival -> immigration/customs -> baggage claim -> exit. Passengers who don't make it through in time are marked MISSED_FLIGHT instead of sitting forever in a stale queue.
• Baggage lifecycle: check-in -> conveyor -> sorting -> screening -> loaded -> unloaded -> carousel -> claimed (or flagged LOST by the AI model before it happens, or OFFLOADED if the passenger missed the flight).
• Gates, runways, taxiways: dynamic assignment/contention, AI-flagged gate conflicts, runway queueing and sequencing.
• Check-in & security: queue-driven throughput, AI recommendations to open/close counters or lanes, congestion alerts.
• Weather Impact Engine: condition drives takeoff/landing probability, taxi speed, and delay accumulation across every active flight.
• Incident simulation + AI Decision Engine: medical emergencies, engine failures, bird strikes, fires, security breaches, fuel shortages, gate conflicts, runway obstructions — each gets an automatic, logged AI response (aircraft reassignment, resource dispatch, runway closure, etc.).
• Resource & ground-vehicle management: staff and vehicles tracked by type/status, surfaced as AI utilization recommendations.
• AI Engine (ai_engine.py): a scikit-learn RandomForestRegressor predicts per-flight delay risk from live weather/queue/gate/maintenance features; a LogisticRegression model predicts per-bag lost-baggage risk before it happens; a rule-based recommender scans every module each tick.
• Digital Twin dashboard (app.py, Streamlit + Plotly): a live aircraft map, full flight/passenger/baggage tables and funnels, gate/runway grids, resource charts, a weather panel, an incident log, and an AI insights page with model feature importance.

Run it

pip install -r requirements.txt
streamlit run app.py

Use the sidebar to advance simulated time (+10 min / +1 hour) and watch every module react together, or reset to a fresh day.

Project layout

• database.py — SQLite schema + connection helpers
• data_generator.py — generates a full synthetic operating day (flights, passengers, baggage, crew, gates, runways, counters, resources, vehicles)
• simulation.py — the tick-based state-machine engine for every module
• ai_engine.py — predictive models + rule-based recommendation engine
• queries.py — read-only helpers that turn DB state into dashboard-ready data
• app.py — the Streamlit dashboard

Notes on calibration

This is a stress-tested single-day simulation (14 flights, ~2,000-2,500 passengers, randomized incidents and weather), tuned so a typical run lands around a 5-10% missed-flight rate, ~3% baggage loss rate, and 85-95% on-time performance under normal conditions — believable numbers for a demo, while still being chaotic enough to showcase the AI/incident layer when weather or congestion spikes hit. Tune the constants at the top of simulation.py and data_generator.py (flight count, throughput caps, incident probability) to make the airport busier/calmer.