VRP on several days

[Alberto Manzini]

Hi everyone, I'm trying to solve a VRP problem in which we want that the vehicles could come back to the depot if needed. For example, in the image below we can imagine that a single vehicle visits all the nodes in 4 different days.I could not figure out how to do it unless using a synthetic vehicle for each day that I want to optimize, then solve each of these fake vehicles and bring the solutions together. In this way, however I have to set the order of the optimized (single) days manually after the solver has finished its job.

Thanks for your help,

Alberto

[haak...@gmail.com]

You have to make nodes representing each depot visit since each node can only be visited once. In your case I would make 4 depot nodes, one for each day. Another alternative is to make a pickup node for every delivery. The pickup nodes can just be copies of the depot, but this allows you to visit the depot again whenever necessary.

[Mizux Seiha]

you should take a look at reload sample:
https://github.com/google/or-tools/blob/stable/ortools/constraint_solver/samples/cvrp_reload.py

Basically add some reload node (or unload node), which are depot duplicate, so the vehicle can visit new locations once full (or empty)...

[Alberto Manzini]

Thanks for help, actually I'm trying to put the capacity refill based on the distance traveled between nodes.

I have started working on https://github.com/google/or-tools/blob/stable/ortools/constraint_solver/samples/cvrp_reload.py as you suggested, simply modifying the "create_demand_evaluator" function, but it's not working.

Someone does have a hint to bugfix this code (see attachment)?

Thanks for your help,

Alberto