Current limitation

We lack a way to better model costs related to using different vehicles. Examples of (potentially) unexpected behavior that can be noticed in output for the sake of reducing overall travel time:

• using more vehicles (Use minimum number of vehicles in solution #495)
• using more expensive vehicles (Vehicles priority #359)

The latter is also now probably more bound to happen as we now have multi-profile in place.

API addition ideas

A way to pass this in the API would be to add a cost object at vehicle level holding a v.cost.fixed value. Then we'd have to provide a way to homogenize with current travel-related costs, maybe with a scaling coefficient like v.cost.per_travelled_hour or similar.

Solving approach

This would require to go through all situations where we modify solutions during the local search phase (all neighboring operators, job removals/additions) to account for change in cost.

The fun part here is that we'd then be able to play with many new HVRP instances.