OpenAP-scikit-decide path optimizer cut A320 fuel burn 3.8% on EWR-FCO using NOAA GRIB winds

Mark's benchmark executed the solver on a Ryzen 9 9950X workstation with 96 GB DDR5. The unified-planning backend generated two candidate trajectories; OpenAP's narrow-body thrust and drag tables converted each into fuel mass. The wind-aware path burned 1,920 kg less Jet-A than the direct route while respecting the same 8.5-hour block time.

OpenAP models were published by Junzi Sun in 2020 and validated against FDR data from 12 European carriers. Scikit-decide added RDDL and Gurobi planners in version 1.0. Jeppesen and Lufthansa Systems offerings use proprietary BADA 4 tables; the open stack replaces them with Sun's polynomial drag curves and NOAA's 0.25-degree RAP fields at zero licensing cost.

Operational effect is immediate for narrow-body fleets. A 3.8% reduction on 1,200 daily A320 sectors yields roughly 180 tonnes of fuel saved per day. Dispatchers can now run the same solver on 4-core laptops inside existing flight-planning cycles without new hardware.

Next release of scikit-decide will expose GPU-accelerated Monte-Carlo rollouts. If OpenAP adds BADA 4 coefficients by mid-2025, the same pipeline will cover wide-body fleets and close the remaining accuracy gap to commercial vendors.