New AI Makes Delivery Routing and Scheduling Up to 5.5x Faster

Every day, businesses face tricky planning questions: How many trucks do you need to deliver packages? How should a hospital schedule surgeries? These are "mixed-integer programming" problems — math puzzles where some answers must be whole numbers. The software that solves them is called a solver, and its speed is critical. If a route planner takes hours, you waste time and money.

Researchers have been using AI to write these formulas automatically from plain English. But the AI often produces messy formulas: completely correct, yet slow. Experts then have to fix them by hand. That's where FormuEvo comes in. Instead of writing a formula once and hoping for the best, it mimics evolution. It generates many formulas, tests each one on a solver, and keeps the best performers. Then it uses the solver's detailed feedback — like "this section caused slowdown" — to create new, better versions. This cycle repeats until the formula runs quickly.

The results are striking. Across diverse problems, FormuEvo's formulas made solvers up to 5.5 times faster than those crafted by human experts or by other AI methods. Even better, the lessons learned on one problem could be reused on another, and smaller AI models benefited too. That means even businesses without massive computing power can adopt this technique.

Why does this matter for you? Faster solving means companies can plan more deliveries, adjust schedules on the fly, and react instantly to disruptions like a closed road. It could mean lower shipping costs, shorter wait times, and less fuel wasted. This is early-stage research, but it points toward a future where AI doesn't just understand your problem — it solves it quickly enough for everyday decisions.