AI-assisted, multi-source cockpit alert prioritization engine designed to reduce pilot workload during high-stakes flight phases.

Modern cockpits can generate dozens of simultaneous alerts from independent systems (TCAS, EGPWS, weather radar, fuel, hydraulics, ATC). Raw alert floods increase crew workload exactly when cognitive capacity is lowest. This project implements a severity-scoring and suppression engine that:

• Fuses alerts from multiple avionics sources in real time
• Scores each alert (0–100) using:
  • alert level base weight (WARNING / CAUTION / ADVISORY)
  • time-to-impact urgency (e.g. TCAS RA time-to-collision)
  • flight-phase relevance (approach amplifies scores)
  • cross-source threat correlation (e.g. TCAS + ATC about the same traffic)
• Ranks all active alerts into a single prioritized stack
• Suppresses low-priority advisories while a WARNING is active, cutting alert noise

The included dashboard runs a scripted flight profile (climb → cruise → descent → approach) with evolving traffic, terrain clearance and weather severity. Alerts are generated from sensor-like triggers, scored, ranked and displayed with a full score breakdown.

• React 19 + TanStack Start (SSR, file-based routing)
• TypeScript — strict typed simulation & scoring engine
• Tailwind CSS v4 — dark cockpit UI

src/lib/simulation.ts   # flight model, alert catalogue, scoring & suppression engine
src/routes/index.tsx    # live dashboard (flight state, alert stack, crew load meter)

score = levelBase + urgency + phaseRelevance + correlationBonus   (capped at 100)

Alerts below threshold are suppressed while any WARNING is active, directly reducing the number of items the crew must process.

bun install
bun run dev

Research/demonstration software — not certified for operational aviation use.