Warm start possibility

[AntoineReot]

Hello, I am currently using the GLNS solver you developed to solve robot task scheduling problems. While the solver performs well and gives us interesting results, I would like to know if it is possible to do a "warm start" by giving to the solver an already computed tour instead of starting from a random one everytime. This should allow for faster computation times and help us a lot as we would like to do several iterations on a given problem. Is it possible to do a warm start or to add such a feature to the already existing code?

[stephenlsmith]

Hi, thanks for the question. I do not believe we currently have this functionality, unfortunately. It would be fairly easy to add since internally the solver does a set of "cold trials" to find a decent initial tour, and then a set of "warm trials" which used the initial tour as a starting point. So, one could read in a tour from file, set it as the "best" tour, set param[:cold_trials] = 0, and then start immediately with the warm trials.

I will add this to a todo, but can't tell you right now when I would be able to add the functionality. If you are interested in working on this, let me know.

Best,
Stephen

[AntoineReot]

Thank you very much for the quick answer. Sadly, I cannot tackle with this myself, as I have no knowledge regarding Julia.

Regards,
Antoine

[negar-abbasi]

Hello, I am currently using the GLNS solver as well, in a robotic area. I need this option to declare the starting point of tour instead of any random starting. I was wondering if you have updated and added this option or not.
Best,
Negar

[stephenlsmith]

No update at this point, unfortunately.