Open ruthmair opened 1 year ago
@torressa @ruthmair could we extend the existing shortest path mod with this version to avoid too much duplication? A simple shortest path problem would still be solvable by imposing no limits.
Indeed a simple shortest path can be used but only for some bounds. This problem has the following additional constraints (as well as others to enforce elementary paths):
$$\sum{ij} a^r{ij} x_{ij} \leq A^r,~~\forall r$$
Where $r$ is a resource (in Mario's example: limits.keys()
). So this would need a change to the underlying min-cost flow formulation (the one in src/gurobi_optimods/network_util.py
)
Yep, I mean we could do away with the min-cost-flow based implementation and only implement the more complex resource-constrained model. If re-using the network_util
stuff over-complicates things, we can just have some duplication.
Ok, then I will take it from here.
@ruthmair @torressa FYI I removed the current shortest path implementation & docs from the main repo, and re-instated it on the branch simonbowly/rcsp, so you can use it as a starting point for docs and tests of a resource-constrained version.
I guess the public/private issue also made the rcsp fork unusable? This is no issue at all, I can just start with a new fork.
Yeah, new fork required I'm afraid, sorry about that.
@ruthmair Are you still working on this? Simon has given us the OK on writing a VRP mod! 😈
I still need to start working on this, but it is on my list. Would you do a VRP mod based on column generation?
Yeah, why not? Would you say the formulation is better? We could have mod based on the formulation(s), or even better formulation(s) + heuristic to insert solutions via callback
Why this Mod?
What will the API be?
Additional context