Route Lab

Fewer kilometres for the same deliveries.

The problem

A van can only carry so much before it has to go back to the depot. With a few dozen customers the number of ways to split them between vehicles is astronomical, and a poor split costs kilometres every single day.

The approach

Route Lab tries three kinds of method on the same customers. Construction methods (nearest neighbour, sweep, savings) build a plan from nothing. Local search then moves and swaps customers between vehicles for as long as that shortens the plan. Simulated annealing goes further and sometimes accepts a worse plan on purpose, to climb out of a rut.

Watching it think

Switch the algorithm and customers change vehicle. Each one can be replayed step by step, so you can see a plan being built or improved, and the plan cost falling.

Try it yourself

Drag customers around, add or remove them, shrink the vans or make every vehicle expensive, and every plan is solved again. On small cases the answers are compared with the exact optimum, which annealing matched every time in my checks.