Routing-solver + use_depth_first_search(true)
-> complete/infeasibility-certificate for initial-solution search?
#4045
Unanswered
sschnug
asked this question in
Routing (and legacy CP) questions
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I'm currently observing the following behaviour:
ROUTING_FAIL_TIMEOUT
=> not able to find solutionuse_depth_first_search = true
with same instance:ROUTING_FAIL
=> not able to find solutionFurthermore we can use
params.mutable_solver_parameters()->set_trace_search(true);
to see:I'm not sure how to interpret this.
I'm inclined to think that CP + DFS without any conflict- or backtracking-limits is always complete as it's full enumeration (supported by propagation).
But in this case, i would expect the return-status to be
ROUTING_INFEASIBLE
. It isn't howewer.(it seems this status is only used in a very limited way:
or-tools/ortools/constraint_solver/routing.cc
Line 2556 in 8768ed7
Does above tell me it's proven to be infeasible and the status is not expressing this or does above not mean, that it's proven to be infeasible?
(as indicated in the verbose-output, we cannot easily use cp-sat or other complete-solvers without being careful about scaling: e.g. not using the assignment-polytope lp-relaxation!)
Beta Was this translation helpful? Give feedback.
All reactions