AeroFlux: From Flight Delay Prediction to a Cyber-Physical Intelligence System

What is AeroFlux?

AeroFlux is, on its surface, a real-time flight delay prediction and explainable analytics platform — streaming SWIM/Kafka ingestion, XGBoost over 61M+ BTS records, and an agent layer that explains why a delay is propagating rather than just scoring it. But the actual thesis is that aviation is a proving ground: flight delay cascades are a measurable, tractable proxy for the general cyber-physical problem of perceiving distributed system state, anticipating downstream consequences, and acting before the decision window closes. The longer arc pushes that same loop outward — cascade simulation and multi-domain orchestration first, then into sensor fusion, robotics, and autonomous systems where the “sensors” are physical and the consequences are kinetic rather than schedule-based. The Swiss Cheese framing holds throughout: the holes are always moving, so the contribution is an intelligence layer that keeps up with them, not a better model.

The Problem It Addresses

Complex systems fail through propagation, not isolated events. In aviation, a single delayed aircraft cascades across crews, gates, and downstream airports — but each actor sees only its own slice of state, and sees it late. The gap isn’t prediction accuracy; it’s the absence of a shared, real-time awareness layer across moving actors operating in a shared environment.

Without that layer, decisions are made on stale, partial state — reacting to disruption after the window to contain it has already closed. The same gap defines robotics fleets, autonomous vehicles, and distributed sensor networks.

What’s Built Today

Stage 1 — Predictive ML Model

A production-scale, propagation-aware ML pipeline trained on 61M+ U.S. domestic flight records (2015–2024) from the Bureau of Transportation Statistics, fused with NOAA hourly weather and airport metadata under strict temporal alignment (no look-ahead bias).

Stage 2 — Real-Time Inference

A live streaming inference and explanation layer that fuses FAA SWIM position feeds through Kafka into a continuously updated state store, scores active flights against the Stage 1 model, and surfaces results through a Dash/Plotly interface backed by a LangGraph agent with pgvector-based retrieval — so the system doesn’t just predict a delay, it explains which upstream conditions are driving it. Deployed end-to-end on AWS Lightsail under Docker Compose at aeroflux.duckdns.org.

Click on hyperlinks above to explore each stage in detail.

Roadmap

AeroFlux is being built in stages, each standing on its own and unlocking the next:

Completed Stages

  1. Stage 1 — Predictive ML Model (complete) — Offline gradient-boosted delay predictor on 61M flights with published methodology and reproducible pipeline.
  2. Stage 2 — Real-Time Inference (complete) — Live data streaming, online serving, agentic copilot, richer feature streams (SWIM, NOTAMs, sentiment).

Future Stages & Exploration

  1. Stage 3 — Network Simulation — Rust-based forward simulator propagating disruptions through rotation chains, gates, crews, and passenger connections.
  2. Stage 4 — Cyber-Physical Intelligence Platform — Operational digital twin of the NAS with pre-tactical decision support and conformance monitoring.
  3. Stage 5 — Multi-Domain Autonomy Layer — Generalize the platform pattern to UAM, drones, AVs, ports, energy, and other systems of interdependent agents in physical space.

Propagate. Simulate. Orchestrate.