Airline Disruption Recovery — Solver Configuration
This repository publishes the solver configuration and benchmark summary for
the Integrated Airline Disruption Recovery System. It is a configuration /
decision-model artifact, not a trained neural network: the "model" is a
multi-stage optimization pipeline combining a time-space network
construction heuristic, Large Neighborhood Search, OR-Tools CP-SAT gate
assignment, and a priority-weighted passenger reaccommodation flow.
Components
| Stage | Method |
|---|
| Aircraft Rotation Recovery | Time-space network + chronological min-cost matching + Large Neighborhood Search |
| Crew Pairing Recovery | Duty/rest legality checker + reserve-pool reassignment |
| Passenger Reaccommodation | Priority-weighted greedy multi-commodity rebooking |
| Gate Assignment | OR-Tools CP-SAT no-overlap interval scheduling |
| Multi-Objective Tradeoff | Pareto frontier sweep: airline cost vs passenger impact |
Cost Model
| Term | Value (USD) |
|---|
| Per minute of flight delay | 45.0 |
| Per flight cancellation | 28,000.0 |
| Per unplanned aircraft swap | 900.0 |
| Per crew pairing reassignment | 650.0 |
| Per misconnected passenger | 340.0 |
See config.json for the full manifest and the latest benchmark summary
across all five demo disruption scenarios.
Intended Use
Decision-support and portfolio demonstration for airline operations control
center (AOCC) style disruption recovery. Not certified for live operational
use; all data is synthetic.
Limitations
- Synthetic hub-and-spoke schedule, not a real airline's data.
- No live weather, ATC, or crew-legality regulatory feed integration.
- Gate assignment and crew legality rules are simplified surrogates of
real FAR/EASA and airport slot rules.
