Hello all,
I
am currently working on a VRP-type problem using chained variables.
Planning entities are tasks to be carried out, and variables are workers
and planned start time to carry out the activity (shadow variable).
It seems beneficial to accomplish my current goals to cluster tasks based on certain properties and plan these clusters, in
addition to the individual tasks themselves. The clusters would have
all the same properties available as an individual task, and therefore
the same rules file should suffice. I would essentially like to plan
the most efficient and fair scheduling of large clusters on a macro
level, while also simultaneously considering the most efficient route
within the clusters. It is a bit like a VRP inside of a job-scheduling
problem. Or a cluster-first route-second type of VRP problem.
I
am currently experimenting with using a custom move to execute the
larger "cluster" moves during local search. I'm not sure if this is the
most efficient way though. I have also read about multiple planned
entities within OptaPlanner, but it is unclear if this would work in the
case of chained variables, or a VRP-type problem.
Does anyone have experience / thoughts here on the best way to accomplish this?
- Andrew