laser-ufpb / cvrpcontroller Goto Github PK

Dear @EduardoQueiroga
Could you clarify if coordinate distances for instance InstancesRounded/F-n135-k7.vrp should be truncated or rounded?
The rules document mentions the below.

All 129 instances from the literature listed for Phase One use EUC 2D distances (but CMT4, CMT5, tai385, and Golden9-20 instances do not follow the TSPLIB95 convention of rounding to the nearest integer).

Kindly clarify.

PS: We should update the README.md. Currently, at two places it points to the older rules document (August 26, 2020).

It looks like the below file Loggi-n601-k42.vrp has NAME field as Loggi-n600-k42 different from the file name.

The controller creates outputs based on NAME field. Because of that, the files were sorted slightly differently when we populated PI values in google sheet. // A very minor thing though, just thought of sharing.

Thanks,
Rajesh.

I understand that we have to drop our cpp solver file in the root folder location and mention the executable name while generating script in arguments like Solver1. Dummy solver prints results in stdout. What about reading input? should we read from stdin or from the argv[1] as input filename?

Hello Team,
When we ran our deterministic heuristic solver (it is not an iterative refinement algorithm) it prints the routes and cost.

build $ ./Solver ../InstancesRounded/X-n101-k25.vrp 1
Route #1: 32 24 31 95
...
Route #30: 41 22 15 5
Cost 33181.167969

When we ran our solver using the controller, it seems to not capture the Solution value, local machine time, standardized time. Though we had done a fflush(stdout) after printing the cost.

./CVRPController Resilience ../InstancesRounded/X-n101-k25.vrp 1 2064 1800 13332 1 Solver
<snip>
Wed 08 Dec 2021 01:05:33 AM IST
timestamp: 1638905733
Solution value, local machine time, standardized time
Primal Integral: 10.0000000000

Will the controller only works for local search algorithms? Do you have any thoughts on this? Are we missing something? Thanks.
Kind regards,
Rajesh