I am currently running a MILP model in a for loop, where I am trying out around 800 different parameter combinations (a kind of sensivity analysis). Sometimes the model doesn’t find a feasible solutions (this is expected) and then interrupts the entire for loop. I am using the Gurobi solver and getting this error: “Result index of attribute MathOptInterface.VariablePrimal(1) out of bounds. There are currently 0 solution(s) in the model.”
Is there any way to just skip to next parameter combination and continue the loop?
I tried implementing this using the code below, but the infeasible configurations were still interrupting the loop.
#handle the error
Thank you for any help in advance!
You could, for instance, only query variable values if you found some result using the result_count() or has_values() function. Something like
if result_count(M) > 0
# query values of interest
@info "no results found"
Alternatively / additionally, you could check for the termination status codes, because infeasibility is not implied when no results were found (which could be also the case if a time limit etc is reached).
Thank you so much for the warm welcome.
I thought the interruption was coming from the infeasible model itself, but you pointed me in the right direction. The interruption was due to calling “value(…)” on an infeasible model.
Following on the idea of
termination_status what I do is something like:
if termination_status(model) == LOCALY_INFEASIBLE
# what you want to do
What I’m looking for now are ways to salvage the optimization cause my application is real-time MPC.
If somebody has any info on this that’d be great.