在OptaPlanner中处理层次约束的推荐方法是什么？

Question

I'm trying to learn OptaPlanner by building a playlist generator. My constraints look roughly like:

1. The total time must be ~60 minutes.
2. All songs selected must be unique.
3. Each ~15 minute block may contain only songs from a single artist.
4. The four 15 minute blocks must be different artists.

So by hierarchical, I mean you could choose the artist and then attempt to fill the block.

My current implementation expresses these as constraints on song selection. It's able to solve the problem, but I feel like it's spending a lot of time trying to align the artist constraint.

Reading through docs, it seems like there are some features which might be helpful:

• Partitioned search
• Chained variables
• Custom move selectors (eg changing all the songs to have consistent artist)
• Different weighting on constraints

What's the recommended way to handle this type of relationship?

Answer 1

Sounds like these are just 4 constraints with 4 different score levels (see docs chapter on score calculation). So in that case, you'll need a BendableScore .

In practice though, I would be surprised if 3. and 4. are hierarchical: 3 might just have a much heavier score weight than 4. In that case a HardMediumSoftScore suffices.