I guess I thought you’d already seen it
I gave that response because I’d read a few articles that day and academia seems determined to get to best-known solutions. It started to irk me I don’t think vehicle routing is used so much these days for that; some companies might want absolute efficiency in planning their fleet movement but actually, for on-demand services, it’s used more as a tool to answer “can I accommodate this job?” and, if so, “when?”. I don’t think new companies are so determined to know that they have the most efficient calculated route (conjecture!). It would be nice if academia and industry could be better aligned on this, but the industry is moving very fast and in divergent ways in the services they provide.
jsprit to be pretty efficient in what it seems to want to be; a general purpose routing algorithm that covers a lot of scenarios (this is a good thing). The paper I linked didn’t have too many constraints. It really depends on how much work you think it would be to implement a new ruin/recreate strategy. The observations about the proportion of jobs ruined on each iteration might be easy enough to implement, I’m not so sure on the ease of the insertion strategy or whether I agree with it in a more constrained problem.