Routing A -> B -> A

[mtygesen]

I have the following duration matrix:

"duration_matrix": [
            [-1, -1, 714, 381, 286, 294], 
            [-1, -1, 507, 174, 120, 127], 
            [719, 537, -1, -1, -1, -1], 
            [406, 224, -1, -1, -1, -1], 
            [283, 142, -1, -1, -1, -1], 
            [428, 247, -1, -1, -1, -1]
   ]

The values "-1" in the duration matrix represent invalid connections.

Conceptually the matrix consists of two sets of vertices: A and B. (duration_matrix[I] < 2 in set A & duration_matrix[I] >= 2 in set B)

I have n vehicles, and I want to route them from some vertice in set A to a vertice in set N, then back to the starting vertice. Is it possible to do this with or-tools? I want to optimize overall duration based on these constraints

[sschnug]

To me it's a bit unclear what you want: maybe add an complete example including a solution.
But it sounds like you should not do any routing at all and just solve an assignment-problem / bipartite-matching problem.

ortools' routing-solver is very generic and might be able to do this but:

• -1's need to be bigMs / int64_t:max() such that they worsen the objective if picked
  • Alternatively / additive: Holes in NextVar domains (underlying CP model of routing-solver) -> domain-filter will disallow moves
• those invalid transitions surely are not that smooth / discontinuous and local-search might suffer

[mtygesen]

Ok so maybe i would be better of just using e.g. the mip solver?