A demonstration shows how open-source tools can shave thousands off long-haul fuel bills by optimizing routes for wind and aircraft performance. A sample calculation estimates a Boeing 787-9 flying Newark to Rome could burn about $68K in jet fuel, and rerouting for winds can cut consumption meaningfully. The workflow couples scikit-decide - a reinforcement learning, planning and scheduling framework - with OpenAP, an aircraft performance and fuel-consumption model developed at TU Delft, and NOAA wind data. The scikit-decide solver can be configured with different fuel models; the write-up compares two A320 flight paths using OpenAP to quantify differences.
Concrete, reproducible setup and data are provided: Python 3.12 in a virtualenv, scikit-decide 1.1.1, openap/opentop, and a list of installed packages (ray, torch-geometric, unified-planning, etc.). Maps were produced with QGIS 4.2.1 and DuckDB (with H3, Lindel, JSON, Parquet, Spatial extensions) for data work. OpenAP bundles extensive aircraft datasets (examples: 17 Boeing, 13 Airbus entries), with rich YAML descriptors - MTOW, wing area, engine options, cruise mach/altitude - and detailed drag polar and phase-specific parameters (cruise speeds, climb/descent rates). The combination of solver, physics-based performance model and meteorological data enables quantitative route optimization and fuel-savings comparisons.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
AI recently solved a version of the Navier-Stokes problem, but this does not mean mathematicians will become obsolete. The machine's achievement addressed a forced version of the problem, leaving the original unforced equations and deeper understanding still reliant on human mathematicians. (hxstem.substack.com)
Jean-Pierre Serre is a French mathematician celebrated for his work in algebraic topology, algebraic geometry, and number theory, turning 100 years old. He developed an early passion for mathematics, studying advanced problems in high school and earning top honors in competitions. (mathshistory.st-andrews.ac.uk)
Quantum computers will not instantly solve hard problems by trying all solutions in parallel, contrary to some misconceptions. AI progress over the past decade has confirmed many wild predictions, indicating that the Singularity has already begun in uneven ways. (scottaaronson.blog)
Medieval mathematicians kept their discoveries secret to secure social and professional advantages, often delaying publication. Today, the culture encourages open publication and sharing, though issues with academic publishing persist. (mathoverflow.net)
A new study suggests Venus may have once had a moon that was eventually pulled in and destroyed by the planet. The models indicate that Venus's early rotation and the moon's mass likely caused the moon to spiral into Venus over time. (nautil.us)
Periodic Labs has developed a high-throughput physical lab infrastructure integrated with AI to accelerate scientific discovery. They trained a trillion-parameter model, Periodic Neon, to analyze experimental results, outperforming some frontier models in scientific tasks. (periodic.com)
Today's best Hacker News stories, summarized and screenshotted, one email a day.